Oxidative damage to DNA constituents by iron-mediated fenton reactions. The deoxycytidine family.

Damage by iron-mediated Fenton reactions under aerobic or anaerobic conditions to deoxycytidine, deoxycytidine-5'-monophosphate, d-CpC, d-CpCpC, and dCMP residues in DNA resulted in at least 26 distinguishable products. Of these, 24 were identified by high performance liquid chromatography retention times, radiolabeling, UV absorption spectra, chemical synthesis, fast atom bombardment mass spectrometry, high resolution fast atom bombardment mass spectrometry, and/or NMR. The nature of the products was qualitatively similar for each substrate except for d-CpC (and possibly d-CpCpC) under anaerobic conditions for which 5-hydroxy-deoxycytidine was uniquely present and 1-carbamoyl-1-carboxy-4-(2-deoxy-beta-D-erythropentofuranosyl) glycinamide was uniquely absent. Damage to dC, d-CpC, and d-CpCpC but not to dCMP or DNA was largely quenched by ethanol, indicating that iron is strongly associated only with dCMP and DNA. The presence of oxygen had little effect with dC or dCMP but had quantitative and qualitative effects with d-CpC and a significantly quantitative but not a qualitative effect with DNA. NADH could drive the Fenton reaction to cause damage to the dC family in vitro, consistent with a previous proposal that NADH was the reducing agent for the Fenton reaction in vivo (Imlay, J.A., and Linn, S. (1988) Science 240, 1302-1309). Finally, the damage spectrum of the dC family by the Fenton reaction is compared with that by ionizing radiation and chemical mechanisms leading to the formation of the 24 identified products are proposed.

damage almost all cell components, including DNA, membranes, and proteins (1). Reduction of H 2 O 2 by reduced transition metals results in the formation of ⅐ OH and related oxidants via the Fenton reaction (2, 3): DNA base damages induced by hypoxanthine/xanthine oxidase (4) or iron and H 2 O 2 (5) have been identified and quantitated. Because iron chelators such as EDTA and buffers such as potassium phosphate were used in these studies, the chemistry of the Fenton reaction has probably been perturbed (6). Such chelators affect the redox potential of iron and may also scavenge oxygen radicals. Moreover, if as is generally believed, the Fenton oxidant is as reactive as is ⅐ OH, then its diffusion distance from DNA is so short that the iron ion involved in the damage is almost certainly complexed to DNA and not to external ligands. Damage to the bases in DNA and mammalian chromatin by H 2 O 2 and transition metals has been investigated (7)(8)(9). However, in these cases chemical hydrolysis procedures were used that might destroy, alter, or form various products (5). In order to gain further insight into the mechanisms of DNA damage by the Fenton reaction, we have therefore chosen to omit exogenous chelators and the use of chemical procedures.
Because killing of Escherichia coli by H 2 O 2 is due to DNA damage and apparently mediated by iron with NADH as the ultimate reducing agent (10 -13), we have also analyzed the degradation of the four DNA base families by iron and H 2 O 2 under a variety of in vitro conditions. In this study of the deoxycytidine family, we have subjected 2Ј-deoxycytidine (dC), 2Ј-deoxycytidine-5Ј-monophosphate (dCMP), d-CpC, d-CpCpC, and DNA to iron/H 2 O 2 under various conditions and then analyzed the resulting dC modifications.
The objectives of this and the accompanying studies were therefore extensively and systematically to study DNA damage without the use of chelators and acid hydrolysis, to identify the major stable degradation products of the dC family after Fenton reactions under a variety of conditions so as to obtain a "fingerprint" of oxidative damages that might be useful for identifying the conditions of Fenton reaction-induced DNA damage from cells subjected to oxidative stress, and to establish the chemical pathways that lead to the formation of the damaged products. Products were identified after enzymatic hydrolysis to the nucleoside level by high performance liquid chromatography (HPLC) 1 retention times, radiolabeling, UV absorption spectra, chemical synthesis, fast atom bombardment mass spectrometry by both positive mode (FAB-MS ϩ ) and negative mode (FAB-MS Ϫ ), high resolution FAB-MS, and nuclear magnetic resonance (NMR). Quantitation utilized either radiolabeling or UV absorption.
The products of ⅐ OH attack mediated by ionizing radiation have been identified for DNA (14 -18) and for cytosine (19 -23). Moreover, dC damage products formed by photosensitization have been investigated by Wagner et al. (24). We therefore also are able to compare damages by these three different agents.
d-CpCpC was synthesized on an automated DNA synthesizer (Applied Biosystems model 380A) using the ␤-cyanoethyl phosphoramidite method, and then the product was purified and the structure was confirmed according to Luo et al. (26).
2-Deoxy-D-ribono-␦-lactone was synthesized from 2-deoxy-D-ribose and bromine. 2-Deoxy-D-ribose (500 mg) was dissolved in 5 ml of doubly distilled H 2 O (ddH 2 O), 60 l of Br 2 was added dropwise, the reaction was incubated on ice for 3 h, and then unreacted Br 2 was extracted with chloroform. The aqueous phase was concentrated under reduced pressure and applied to a home-packed silica HPLC column (250 ϫ 4.6 mm), which was equilibrated with 75% ethyl acetate-16% iso-propanol-9% ddH 2 O. The eluate was concentrated under reduced pressure, redissolved in ddH 2 O and purified on a C 18 RP HPLC column for which the mobile phase was ddH 2 O. The identity of 2-deoxy-ribono-␦-lactone was confirmed by its NMR spectrum and by FAB-MS ϩ .
Oxygen Removal and Oxygen Concentration Measurements-The oxygen concentration in solution and the rate of O 2 consumption due to the Fenton reaction were measured with an Orion Research oxygen electrode model 97-08. Air was removed from reaction solutions by sparging with N 2 for 20 min prior to the addition of H 2 O 2 and then continued sparging throughout the 30-min incubation period. Within 3 min of N 2 sparging, the pO 2 was below detectable levels (less than 1% of ambient).
Fenton Reaction Conditions-Substrates were exposed to Fe 2ϩ /H 2 O 2 under anaerobic conditions and to Fe 2ϩ /H 2 O 2 , Fe 2ϩ /NADH/H 2 O 2 , or Fe 3ϩ /NADH/H 2 O 2 , with or without ethanol under aerobic conditions. The damage products (except where dC was substrate) were digested to nucleosides with DNase I, snake venom phosphodiesterase, P1 nuclease, and/or bacterial alkaline phosphatase as appropriate, separated first by normal phase silica HPLC and then purified by HPLC and analyzed, all as outlined previously (26). In spite of the high resolving power of normal phase (silica) HPLC for dC degradation products, this method was not applicable for analytical studies because on-line quantitation and identification by UV absorption spectroscopy was interfered with by the presence of UV-absorbing ethyl acetate in the mobile phase. Refractometry (24) was also not amenable to quantitation because the refractive indices of standard solutions of many of the products are not known. Therefore RP HPLC was utilized for separation of the products, and UV absorption was used for on-line quantitation and preliminary identification.
FAB-MS and NMR-For positive ion mode FAB-MS, samples from dC or dCMP reactions were mixed with a glycerol matrix, which also served as an internal marker and then applied to the probe tip for bombardment by a stream of xenon gas in a Kratos Mass Spectrometer model 50. For negative ion mode FAB-MS, samples from dCMP or d-CpC reactions without enzymatic digestion were dried under reduced pressure and subjected to FAB-MS Ϫ with glycerol added as the matrix and internal marker. For NMR spectroscopy, purified samples were dried and redissolved in 5 ml of D 2 O, and spectra were recorded on an AM400 NMR spectrometer.
Other Methods-Preparation of unlabeled or dC-radiolabeled phage PM2 DNA, enzyme assays, determination of concentrations of hydrogen peroxide and iron, and HPLC were described previously (26). The concentration of PM2 DNA was estimated by UV absorption, assuming that ⑀ 260 ϭ 6.5 mM Ϫ1 cm Ϫ1 . dC and (poly)nucleotide concentrations were also determined by UV absorption.

Establishment of Reaction Conditions
2 mM H 2 O 2 was used for Fenton reactions because this concentration gave maximal killing of E. coli (12) and should reflect both Mode I and Mode II conditions (27). Moreover, it is biologically more relevant than very high H 2 O 2 concentrations or H 2 O 2 -generating systems, which are somewhat difficult to define, both of which are often used. Because Fe 2ϩ was limiting (except with NADH), the concentration of iron determined the amount of damage, and, for example, the extent of damage to dCMP was proportional to ferrous concentrations up to 1 mM (data not shown). In general, the iron concentration was the minimum necessary for sufficient product characterization, but secondary damages were statistically possible under these conditions if total damage exceeded 10 -20%.
To investigate the effect of pH, 1 mM dCMP was reacted with 1 mM Fe 2ϩ and 2 mM H 2 O 2 from pH 4 to pH 8 ( Fig. 1). dCMP was chosen for this investigation because of its inherent buffering capacity. As expected for the Fenton reaction, it was more efficient at low pH, but there was a transition between pH 6 and 7. Conversely, the formation of free cytosine, one of the major dCMP damage products, increased with increasing pH (Fig. 1), perhaps indicating that high pH stabilizes an intermediate that leads to the release of cytosine. The effects of pH upon dCMP reactivity might have reflected: 1) changes of the binding of Fe 2ϩ due to changes of the charge on dCMP (the secondary pK a of dCMP-phosphate is 6.6) (28); 2) the difference of reactivity of protonated dC and its analogs to that of unprotonated forms (the pK a 1 of the cytosine ring is 4.3); 3) the enhanced oxidation of ferrous by oxygen above pH 6.5 (29). Parenthetically, the retention times of many dC-derived products on C 18 RP HPLC columns were sensitive to pH of the elution buffers in the range 4.5-7, presumably also because they have pK a values in this range. pH 6.5 was found to give the best resolution on C 18 RP HPLC columns.
Based on the above trials, 1 mM substrate was reacted with 2 mM H 2 O 2 and 0.4 -1 mM iron at pH 6.5. pH 6.5 was chosen  4 were added to dCMP that had been titrated with dilute NaOH to the pH values indicated. After 30 min at 25°C, the mixture was treated with bacterial alkaline phosphatase and then injected onto a C 18 RP HPLC column for separation and analysis as described under "Experimental Procedures." The amount of dCMP damaged was calculated from the disappearance of dC in the HPLC elution profile. 0.03 mM cytosine was produced at pH 4.0. Similar results were found when thin layer chromatography was used to monitor the disappearance of dC.
because it is close to physiological pH, because ferrous is easily oxidized by oxygen at or above pH 7 (30,31), and because the pH of 1 mM solutions of dC were 6.5-6.8 so that exogenous buffers that might alter the redox potential of iron, scavenge Fenton oxidants, and/or compete for iron binding could be avoided. FeCl 3 solutions were freshly prepared and were free of precipitate, perhaps because Fe 3ϩ was still in the [Fe(H 2 O) 6 ] 3ϩ form. (The concentration of free Fe 3ϩ at equilibrium with an insoluble bridged precipitate at physiological pH is estimated to be 10 Ϫ18 M (32).) No precipitate was observed following the Fenton reactions, presumably because the ferric ions were chelated by the substrates and their damaged derivatives (33).
Characterization of the Damages dC as Substrate-Systematic studies were carried out with either unlabeled dC (not shown), [1Ј,2Ј,5-3 H]dC ( Fig. 2A), [5-3 H]dC (not shown), or [U-14 C]dC (not shown), and the damaged products were traced by radiochromatograms and UV absorbance between 190 and 430 nm (26). Most of the products absorb maximally at wavelengths less than 230 nm, whereas dC has an absorption maximum at 271 nm.
The identity of the products was also determined by FAB-MS ϩ , high resolution FAB-MS ϩ , NMR, radiolabeling, chemical synthesis, UV spectra, and HPLC retention times based on standard products that were either obtained commercially or synthesized (Table I) (26).
1-Carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was the major dC-derived product from aerobic Fenton reactions. Based on the radioactivity to absorbance ratios for the three radioactive substrates, it is concluded that this product retains all of the 1Ј, 2Ј and C-5 tritium labels, indicating that all or part of both the sugar and base moieties are still present. The UV absorption spectrum indicates that the aromaticity is lost from this compound, so the base must be damaged, whereas the sugar is likely to be intact. Its structure was determined by its 1 H-NMR (D 2 O) spectrum (Fig. 3A), FAB-MS ϩ (Fig. 4) and high resolution FAB-MS ϩ for which the masses obtained were 278 and 278.098 by low and high resolution, respectively, (versus 278.2419 as calculated for C 9 H 16 N 3 O 7 ). Smaller mass spectrometry peaks are consistent with fragments of this product (see legend, Fig. 4). 1-Carbamoyl-1-carboxy-4-(2-deoxy-␤-Derythropentofuranosyl)-glycinamide absorbs well below 205 nm with a maximum at 194 nm. Although this product is the major product from dC in the aerobic Fenton reaction, it is only a minor product in type I photosensitization of dC (24) and Fenton reactions with DNA (see below).
Slightly more damage to dC by Fe 2ϩ and H 2 O 2 was observed under anaerobic conditions (Table II), and the product distribution was somewhat altered (data not shown), but in general oxygen has little effect upon damage to dC (Table II). at pH 6.5 aerobically. After 30 min at 25°C, DNase I, nuclease P1, snake venom phosphodiesterase, and bacterial alkaline phosphatase were added to convert DNA or damaged DNA to nucleosides as described under "Experimental Procedures," and the sample was then injected onto a C 18 RP HPLC column for separation and analysis. All radioactivity profiles were obtained by continuously collecting small samples and determining their radioactivity. Numbers correspond to the products listed in Table I. Ethanol almost completely quenched damage to dC (Table  II), indicating that ferrous ion was probably not intimately associated with dC, thus allowing the Fenton oxidant to be scavenged by the ethanol. Damage to dC by Fe 2ϩ /NADH/H 2 O 2 was also sensitive to ethanol (Table II), a surprising observation, because damage to NADH in the same reaction was resistant to ethanol. Evidently some fraction of Fe 2ϩ is intimately associated with NADH (but not with dC) so that ethanolresistant radicals that damage NADH but not dC can form. Similar ethanol inhibition with Fe 3ϩ /NADH/H 2 O 2 was observed (Table II).
In all, at least 26 products were produced from Fenton reac-tions with dC under aerobic conditions, the majority of which were derived from modification of the base (Table I). The possibility that some of these "cytosine-derived" products might have been artifacts obtained from FAB-MS was excluded because parabanic acid, for example, was detected even before FAB-MS. The nature of the dC degradation products was generally similar under the various conditions of exposure to H 2 O 2 and iron, though the yields and the product distributions varied (data not shown). All of the dC-derived products except 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)glycinamide have also been reported after ionizing radiation of cytosine (where appropriate) and dC (20, 36 -38), and some  were found after photosensitization (24,35). 2 dCMP as Substrate-Studies with dCMP are of interest to determine whether perturbations of iron binding by the phosphomonoester alter the product spectrum from that of dC. After the Fenton reaction, the products (and residual dCMP) were dephosphorylated with bacterial alkaline phosphatase, and then the mixture was injected onto a C 18 RP HPLC column for separation and analysis as was used for dC.
With Fe 2ϩ /H 2 O 2 under aerobic conditions, dCMP incurred the same amount of damage as dC ( Fig. 2B and Table II). The nature of the products from this dCMP reaction was essentially the same as that from dC based on FAB-MS ϩ and the similarity of their elution profiles on RP HPLC. In contrast to the dC Fenton reactions, however, 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was only a minor product (compare peak 25 in Fig. 2, A and B).
The amount of damage to dCMP by Fe 2ϩ /H 2 O 2 under anaerobic conditions was only slightly enhanced compared with that under aerobic conditions (Table II), indicating that in this case, oxygen does not play a major, quantitative role in the extent of reaction. However, as opposed to the case with dC, the majority (65%) of dCMP damage was not quenchable by 100 mM ethanol (Table II), indicating either that the oxidants were different than those formed with dC or that they were intimately associated with dCMP when formed. Similarly, with Fe 2ϩ /NADH/ H 2 O 2 , although the amount of damage was roughly the same without ethanol, only about 10% of dCMP damage was prevented by 100 mM ethanol (versus almost total quenching with dC) (Table II). Again, the phosphate of dCMP affects the types of radical formed or their juxtaposition to substrate even with NADH present. However, Fe 3ϩ /NADH/H 2 O 2 caused less damage to dCMP than to dC (Table II), indicating that ferric ions may have been bound to dCMP and hence not so effectively reduced by NADH.
In general, differences in relative product yields and ethanol inhibition were observed under all conditions; however, there was no variation in the types of damage produced (Table II and data not shown). d-CpC as Substrate-Following the same strategies as above, d-CpC degradation was systematically investigated.  (Fig. 2C) was more akin to that of dC ( Fig. 2A) than to that of dCMP (Fig. 2B). For example, 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was one of the major products from d-CpC. There were, however, two major differences between the reactions of d-CpC and of dC with Fe 2ϩ /H 2 O 2 under aerobic versus anaerobic conditions: 1) about twice the amount of d-CpC was damaged under anaerobic conditions compared with aerobic conditions (Table II), whereas there was no such difference with dC; 2) under anaerobic conditions, 1-carbamoyl-1carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was not produced, whereas 5-hydroxy-deoxycytidine (5-OH-dC) was produced immediately by the Fenton reaction, unlike the case of dC, for which 5-OH-dC was observed only after 2 weeks of storage at Ϫ20°C.
Like the case of dC but unlike that of dCMP, damage to d-CpC was severely quenched by ethanol (Table II). This observation suggests that the phosphodiester in d-CpC behaves very differently from the phosphomonoester of dCMP, probably due both to charge differences and steric hindrance. Fe 2ϩ ion may be weakly associated with the phosphodiester group in d-CpC molecule, or alternatively, it might associate with d-CpC at the phosphodiester group and with one of the two N-3 atoms (or just the two N-3 atoms).
Like the case with dC, d-CpC damage caused by Fe 2ϩ / NADH/H 2 O 2 and Fe 3ϩ /NADH/H 2 O 2 was also severely quenched by the presence of ethanol (Table II). Interestingly, the formation of 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was preferred with Fe 3ϩ / NADH, as was the case for dC. The chemical basis of this phenomenon is unknown. In general, the nature of the d-CpC degradation products was the same as with dC under different reaction conditions with the exception of anaerobic reactions, where a novel product, 5-OH-dC, was produced, whereas 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was not formed. Damage by Fenton reactions to the 3Ј-and 5Ј-dT of d-TpT was found to be indistinguishable, 3 so it might be expected that the same would hold for the two nucleosides of d-CpC. However, such an investigation with the 5Ј-nucleoside of CpC specifically labeled was not done.
d-CpCpC as Substrate-d-CpCpC was initially thought to be a better model for DNA damage than dCMP or d-CpC. Surprisingly, however, the damage of d-CpCpC was quite different from that of DNA (see below) based on HPLC elution profiles and UV spectra analyses (data not shown). With Fe 2ϩ /H 2 O 2 under aerobic conditions, the only product definitely identified from UV absorption chromatograms was unaltered cytosine base. Peaks at positions corresponding to dR-formamide, trans-1-carbamoyl-imidazolidone-4,5-diol 2-deoxy-D-ribono-␦-lactone, and possibly 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide were the remaining major products of d-CpCpC under aerobic conditions. Unlike the case for DNA (see below), oxygen did not play a major role in the damage of d-CpCpC, and a significant amount of d-CpCpC damage was quenched by ethanol (Table II)

. FAB-MS ؉ of 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide.
A peak at 176 (fragmented base attached to deoxyribose) has a formula of C 7 H 14 NO 4 (calculated, 176.1925; measured by high resolution, 176.0923); peak 162 (fragmented base only) has a formula of C 4 H 8 N 3 O 4 (calculated, 162.1252; measured, 162.0518); peak 117 is presumably decomposed deoxyribose ion, which has a formula of C 5 H 9 O 3 ; peak 105 is presumably damaged base, which has a formula of C 2 H 4 O 3 N 2 . Peak 197 is presumably an adduct of 105 with glycerol. The above four peaks presumably are derived from the molecular ion of 278. Peaks 131, 153, 185, and 223 are matrix (glycerol) or its adducts with Na ϩ or K ϩ ions. The spectrum was recorded on a Kratos Mass Spectrometer model 50. H]dC-labeled PM2 DNA were used as substrates. The extents of damage were calculated from radioactivity recovered in the dC HPLC peak, except for the d-CpCpC reactions (numbers in parentheses), which were based on UV absorbance. 0.4 mM iron was used when dC, dCMP, and d-CpCpC served as the substrate, whereas 1.0 mM iron was used with d-CpC* and PM2 DNA. 1 mM dC, dCMP, d-CpC (molecules), 1 mM PM2 DNA (nucleotide residues), or 0.33 mM d-CpCpC (molecules) were present as indicated. NADH at molar concentrations that were twice those of iron and/or 100 mM ethanol were added as indicated. Reactions are described under "Experimental Procedures." EtOH, ethanol; ND, not done. Means and standard deviations are given for experiments repeated several times.  (40). With this substrate, about 9 and 14% of the dC moieties were damaged under aerobic and under an anaerobic reactions with Fe 2ϩ /H 2 O 2 , respectively (Fig. 2E, Table II), a level similar to that with d-CpC. However, 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was not a major product from DNA, unlike the case of d-CpC, where it was. The majority of damage to the dC moiety of duplex DNA was not ethanol-quenchable, as is the case for DNA nicking (27).
In the presence of NADH, much more damage to the dC moiety in DNA occurred than to dC, dCMP, d-CpC, or d-CpCpC. There was a more than two-fold increase with Fe 2ϩ / NADH/H 2 O 2 than in the absence of NADH, and only 14% of the damage was ethanol-quenchable (Table II). This observation strongly suggests the formation of a ternary complex among DNA, Fe 2ϩ , and NADH, which is possibly unique to the duplex DNA. In order to test whether such a complex might exist, equilibrium dialysis under anaerobic conditions was used to measure the amount of Fe 2ϩ associated with DNA (the excess inside of the dialysis bag) with or without NADH present. Indeed, a higher amount of Fe 2ϩ was associated with DNA when NADH was included (Table III), supporting the hypothesis that a complex forms among DNA, Fe 2ϩ , and NADH.
The enhancement effect of DNA damage by NADH was also observed with Fe 3ϩ /H 2 O 2 . DNA damage was 50% higher with Fe 3ϩ /NADH/H 2 O 2 than with Fe 2ϩ /H 2 O 2 , and ethanol quenched this damage by about 30% (Table II). Therefore, a complex is probably also formed among DNA, Fe 3ϩ , and NADH.

Comparison of Damages among the Various Substrates-
The nature of the dC degradation products from DNA were basically the same as that from dC, dCMP, and d-CpC based on the analysis of their elution profiles and UV spectra, although the yields and the product distributions differed somewhat. However, dC, dCMP, d-CpC, nor d-CpCpC was a perfect model for DNA because the damage enhancement in the presence of NADH is unique to duplex DNA. Various comparisons of the degree of damage to the members of the dC family after Fenton reactions under the various conditions are summarized in Table IV.
Effects of Oxygen-Oxygen is a significant factor in damage by ionizing radiation of constituents of nucleic acids (41); however, it did not appear to affect the amount of damage to dC and dCMP caused by Fenton reactions, although slightly enhanced damage occurred under anaerobic conditions. Because oxygen is produced as an intermediate in the Haber-Weiss cycle (2), it   is possible either that the oxygen produced from the decomposition of H 2 O 2 was not completely removed by N 2 flushing or that H 2 O 2 can substitute for oxygen in reacting with carboncentered radicals. Alternatively, oxygen may not directly take part in the damage of these substrates during Fenton reactions.
In order to distinguish among these hypotheses, the oxygen concentration was measured during the Fenton reaction of Fe 2ϩ with H 2 O 2 under aerobic conditions. In the absence of substrate, the Fenton reaction produced oxygen as expected (Table V); however, the oxygen concentration decreased when substrates were present, and a more rapid decrease in the oxygen concentration occurred with dC or d-CpC than with dCMP (Table V), although dC and dCMP were damaged to the same extent (Table II). This result may indicate that the dCMP radicals formed are not accessible to oxygen so that less oxygen FIG. 6. Proposed pathways that lead to the formation of all of the products detected with the deoxycytidine family. The pathways that lead to the production of dC degradation products might involve at least seven different types of reactions: Panel A, 1, attack by radicals similar to ⅐ OH on dC at C-5, C-6, or C-1Ј lead to the formation of a dC radical, dC ⅐ . 2, the formation of O 2 adducts, dC-O-O ⅐ , occurs by reactions between O 2 and dC ⅐ . The products of 2 may in turn undergo either 3, intramolecular rearrangement to form products 21, 24, and 25, or 4, deamination to form a variety of products, such as products 1, 14 and 16. 5, iron may also be involved in the formation of and/or reaction with dC-OOH intermediate, leading to the formation of products 5, 9, 17, and 23. 6, secondary attack on dC ⅐ by another ⅐ OH-type radical may lead to the formation of product 6. 7, dC ⅐ radical may rearrange to form 2-deoxy-D-ribono-␦-lactone or cytosine. Panel B, secondary attack on cytosine by ⅐ OH may lead to the formation of products 2, 4, 10, 12, 13, 15, 19, 20, and 22. Alternatively, these products may be derived from the damaged bases, whose intramolecular rearrangement then results in separation from the sugar moieties, thus leading to the formation of these products. The numbers in parentheses correspond to those in Table  I. O 2 remains present under aerobic conditions due to its generation from peroxide (Table V). is consumed. The observation that more oxygen is consumed by dC or d-CpC is consistent with an oxygen requirement for the formation of 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide, because it is the major product from both dC and d-CpC but only a minor one from dCMP aerobic Fenton reactions (see "Discussion"). DISCUSSION dC is the most sensitive of the four deoxynucleosides to aerobic Fenton reactions. 70% of the dC is damaged with 1 mM Fe 2ϩ /2 mM H 2 O 2 present, whereas 55% of dT, 31% of dA, and 28% of dG are damaged (39) (Fig. 2E), compared with 12% of the dT, 6% of the dA, and 8% of the dG under the same conditions (39). 3 Therefore the greater sensitivity of dC and dCMP to iron/H 2 O 2 compared with the other three base families is not reflected in the dC moiety in DNA.
It is clear from the product profiles (Table I) and previous studies (17,24) that the main attack site of oxidative radicals on dC is the C-5-C-6 bond. Based on quantum chemical considerations, Pullman and Pullman (42) predicted that the C-5-C-6 bond has the highest electron-donating capacity within cytosine or cytosine nucleosides, leading Van de Vorst and Westhof (43) also to argue that the C-5-C-6 bond is the most likely target of oxidative radicals such as the OH radical. From the present study it also appears that the C-5-C-6 bond is the major site of attack for radicals generated by the Fenton reaction.
It appears that 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide is produced mainly by freely diffusible ⅐ OH or a similar radical because 1) it was the major dC-derived product ( Fig. 2A) and such a diffusible radical is presumably the damaging radical in these reactions; 2) it was a minor product with dCMP (Fig. 2B), for which freely diffusible ⅐ OH is apparently not the major damaging species. Moreover, formation of 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide was not observed under anaerobic reactions with d-CpC, suggesting that oxygen is required for the formation of this product.
The damage to d-CpC and DNA was enhanced by at least 50% under anaerobic versus aerobic conditions, and in addition, 5-OH-dC was produced only with anaerobic d-CpC Fenton reactions. Conversely, 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-Derythropentofuranosyl)-glycinamide was not detected under these conditions. Oxygen may participate in a way to enhance formation of some degradation products (e.g. 1-carbamoyl-1carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide) or to inhibit formation of others (e.g. 5-OH-dC formation from d-CpC). One plausible mechanism for the oxygen inhibition is that under anaerobic conditions, a carbon-centered radical on DNA or d-CpC is formed due to the Fenton reaction and the nascent Fe 3ϩ or another Fe 3ϩ ion reacts with this carboncentered radical to regenerate Fe 2ϩ ion, which then undergoes a second cycle of Fenton reaction to cause further damage to DNA or d-CpC. However, if O 2 were present, it could react with the carbon-centered radical and a peroxy radical R-OO ⅐ would form, which would then lead to the formation of different products, such as 1-carbamoyl-1-carboxy-4-(2-deoxy-␤-D-erythropentofuranosyl)-glycinamide. The peroxy radical might also oxidize Fe 2ϩ , thus leading to less overall damage. A scheme that summarizes these suggestions is shown in Fig. 5. Why the formation of 5-OH-dC would be substrate-dependent is unclear, but it might reflect a unique type of complex with iron ions.
Equilibrium dialysis indicated that a ternary complex is formed among DNA, Fe 2ϩ , and NADH. Such a complex may also exist among DNA, Fe 3ϩ , and NADH, although it is almost impossible to directly determine the equilibrium binding constants for DNA, Fe 3ϩ , and NADH because Fe 3ϩ will be reduced by NADH before an equilibrium is established. The enhanced damage (compared with Fe 2ϩ /H 2 O 2 ) due to the presence of NADH occurred only with DNA (see Table II and Table V), not with dC, dCMP, d-CpC, and d-CpCpC, suggesting that the binding of NADH (and iron) to DNA allows NADH to efficiently drive the Fenton reaction to cause damage. This observation suggests that an NADH-iron complex might intercalate into DNA or bind to one of the DNA grooves through both electrostatic and van der Waals' interactions. However, the NADH/ iron complex cannot intercalate into d-CpC or d-CpCpC because these substrates cannot provide efficient base stacking and do not contain stable secondary structures.
A comprehensive scheme of alternative pathways that lead to the formation of each of the products detected with the dC family is proposed in Fig. 6. This scheme is obviously tentative and somewhat speculative, so it should be taken as a working model for future experiments. Measurement of the products found in the scheme might be used as an index of oxidative attack upon DNA in living tissues or in isolated cells (44). Moreover, comparisons of the relative amounts of some of the products might be useful to discern the in vivo reaction environments under different conditions of oxidative stress, a final goal of these studies.