Guanine Oxidation in Double-stranded DNA by MnTMPyP/KHSO5: At Least Three Independent Reaction Pathways

In order to better define the mechanism and the products of guanine oxidation within DNA, we investigated the details of the mechanism of guanine oxidation by a metalloporphyrin, Mn-TMPyP, associated to KHSO5 on oligonucleotides. We found that the three major products of guanine oxidation are formed by independent reaction routes. The oxidized guanidinohydantoin (1) and the proposed spiro compound 3 derivatives are not precursors of imidazolone lesion (Iz). These guanine lesions as well as their degradation products, may account for non-detected guanine oxidation products on oxidatively damaged DNA.


INTRODUCTION
DNA is a critical cellular target for several oxidation reactions which are associated with aerobic cellular metabolism or to exposure to physical and chemical agents. Oxidative DNA damage plays a key role in mutagenesis, carcinogenesis and cellular aging. Reaction of hydroxyl and alkoxyl radicals, singlet oxygen as well as one-electron oxidants were shown to be responsible for DNA damage, x'2 Among the four nucleobases, guanine is the preferred target because of its high electronic density and its lowest oxidation potential. However, the nature of the oxidation products and the mechanism of their formation is not well established. One of the most common oxidative modification at guanines is 8-oxo-7,8-dihydro-guanine (8-oxo-G) which is detected in several oxidized DNAs. 4'5 The formation of 8-0xo-G by the attack of reactive oxygen species like OH or 102 is well documented. 46 However, its formation in the case of electron tranfer mnor reaction athwa 0 A second well known roduct of atone oxidation s oxidation may represent a P Y" P gu imidazolone (lz) r'13 associated to its hydrolysis product oxazolone (Z) This product appears as the major one under electron transfer oxidation conditions on oligonucleotides. 7'9'1 A rigorous quantification of Iz and 8-0xo-G was undertaken in the case of oxidation of calf thymus DNA with riboflavin known to be mainly a type I photosensitizer (a one-electron oxidant). The amount of these two oxidation products was inferior to the quantity of the total oxidized guanine. In the case of oligonucleotides this difference was even higher. This data implies that some products of guanine oxidation, awaiting identification, may be of significant 1415 importance. Because of its low oxidation potential (lower than the one of G) 8-oxo-G was pstulated to 101 be a transient intermediate in the mechanism of guanine oxidation through electron transfer.
Thus, some o/ther products resulting from 8-oxo-G oxidation may account for the observed missing material. The oxidation of 8-oxo-G derivatives (in the case of an electron transfer mechanism) led indeed to the characterization of new oxidation products, namely guanidinohydantoirL a spiroiminohydantoin and oxaluric acid derivatives, a618 However, 8-oxo-G is not a necessary intermediate in the reaction of guanine oxidation since we reported that the cationic manganese porphyrin (Mn-TMPyp)9 associated to an oxygen atom donor SO5) was able to perform the quantitative transformation of 2'-deox3'guanosine into imidazolone nucleoside (dlz) without intermediate fomaation of 8-oxo-G. 2 The oxidation of guanine in double-stranded (ds) oligonucleotides (ODNs) by the same system, Mn-TMPyP/KHSOs, 2 did not lead to Iz only, but to three main products, a dehydro-guanidinohydantoin (1) that was the major product of guanine oxidation in this case, imidazolone (Iz) and a proposed 5,8-dihydroxy-7,8-dihydroguanine (2) (Figure 1). No 8-oxo-G was detected but new oxidation products. These new guanine oxidation products could therefore be considered as good candidates for the missing material previously observed in guanine oxidation on DNA. Two of the guanine oxidation products, the imidazolone derivative (Iz) and the new compound, oxidized guanidinohydantoin (1) derivative, were characterized by NMR spectroscopy through the oxidation of a dinucleoside monophosphate model, d(GpT), by Mn-TMPyP/KHSO.. 22 The new compound proposed to be 5,8-dihydroxy-7,8-dihydroguanine (2) on the basis of ESI-MS data has not been further characterised up to now. From literature data, the structure 2 may be unstable and the compound may rather correspond to a spiro derivative with the same molecular mass (3) t6'23"25 ( Figure 1).  Figure 1: Structure of the products of guanine oxidation by electron transfer mechanism. Their molecular mass is indicated with respect to that of guanine, dR stands for 2-deoxyribose or an intmstrand linkage.
In order to better define the mechanism and the products of guanine oxidation, we investigated the details of the mechanism of guanine oxidation by Mn-TMPyP/KHSO5 on oligonucleotides.
We found that the three observed products of guanine oxidation in ds DNA are formed by independent reaction routes. The oxidized guanidinohydantoin (1) and the proposed compound 3 derivatives are not precursors of imidazolone lesion (Iz). We also studied their stabilities. The oxidation products as well as their degradation products do not show high UV absorption coefficients. They, or their degradation products, may account for non-detected guanine oxidation products on oxidatively damaged DNA.

MATERIALS AND METHODS
Materials. Potassium monopersulfate, KHSO5 (triple salt 2 KHSOs. K2SO4. KHSO4, Curox(R)) was from Interox. Oligonucleotides were synthesized by standard solid-phase [3-cyanoethyl phosphoramidite chemistry. They were purified by HPLC using a semi-preparative reverse-phase column (Nucleosil C18, 10 tm from Interchrom, Montluqon, France); eluents, A 0.1 M triethylammomum acetate (TEAA) (pH 6.5), B CH3CN; linear gradient, 1% to 20% B over 60 min; flow rate, 2_.5mL/min; 260 nm). Mn-TMPyP was prepared as previously described. 6  Stabili.ty of lesions I and 3. ODN I was oxidized on a two fold scale compared to the typical reaction conditions described above. The collected fractions of oxidized strand containing one lesion 1 or one lesion 3 were lyophilized and then dissolved in 200 txL of water separately. The purity., of the collected product was analyzed by HPLC/ESI-MS on half of the solution (100 ptL) while the other 100 lxL sample was heated at 90 C for 15 nfin and then analyzed by HPLC/ESI.MS.
Evolution of lesion 3 in water andpH 8.8. The oligonucleotide strand containing the spirt G+34 lesion 3 was collected from a 10 fold scale oxidation of ODN I. After the lyophilization procedure, half of the material was recovered in water (320 L) while the other half in 320 tL of Tris/HC1 buffer (pH 8.8, 100 mM). Each solution was then divided in four aliquols. One of them was analyzed by HPLC immediately while the other three were heated at 90 C in a water bath. for 15, 30 and 60 min respectively after which they were analyzed by HPLC. In the same manner, the stability of purified spirt G+34 was monitored in Tris/HC1 buffer (pH 7, 50 raM). In a separate elrimenl, an aliquot of spirt G+34 was heated in water at 90 C for 60 rain. After which Tris buffer at pH 8.8 (final concentration 100 mM) was added and the mix-ture was heated at 90 C for 15 rain and anyzed by LC.

RESULTS AND DISCUSSION
We previously reported that the DNA degradation mediated by Mn-TMPyP/KHSO5 within G-rich ds ODNs, were due to chemical modifications of guanines. The Mn-TMPyP/KHSO5 sy.stem used for the oxidation of double-stranded G-rich oligonucleotides has the advantage ofbeing a pure electron abstracting agent (no traces of O nor O2 were formed during the reaction). The mechanism was shown not to be a "one-electron" tranfer mechanism but a "two-electron" oxidation of guanine with the intervening of a guanine-cation (G-H)' irtstead of a guanine radical (G-H) ..0 Labeling experiments in H80 allowed us to propose a mechanism of guanine oxidation which involved the formation of 2 by trapping of the guanine cation at C5 by a moleofle of water followed by the attack of another molecule of water at C8. This product was then proposed to be oxidized nto I to give k. (see Figure 1 for stnctures). Compound I was also proposed to result from the attack of HSOf at C5 of the guanine cation. 2 In the first part of this work we investigated guanine oxidation by .Mn-'ITVIPyP/KHSO5 on two different oligonucleotide sequences in order to verify that the product distribution was not sequence dependent. However, the single-over doublestranded structure of DNA was found to influence the distribution of the guanine oxidation products. In the second part, the 80-labeiing of the products wa carefully measured to confirm the competition between H:O and HSOs as trapping nuleophiles on the route of the formation of the oxidation products. Finally, the stability of the oxidation producs wa tudied.
1-Oxidation of guanine in single-and double stranded oligonucleotides by Mn-TMPyP/KHSO The G-rich double-stranded duplexes selected in tiffs work consisted ofthe self complementary 5'-CAGCTG (ODN I) for the first one and 5'-CAGGTG (ODN IIa) hybridized with 5'-CACCrG (ODN IIb) for the second one. The oxidation reactions were carried out in 50 mM Tris/HC1 buffer pH 7, 100 mM NaC1, at 0 C for 5 mix using a ODN (duplex)/Mn-TMtP/KHSO5 ratio of 1:1:50, the concentration of the duplex ODN was 10 tM. The analysis was performed using reverse phase HPLC coupled with negative electrospray mass stxtrometff (ESI-MS). The concentrations of the reactants were significantly lower than the ones used previously due to the on-line turbo ion spray source that increased the detection sensitivity. The liquid chromatography allowed the sodium counter ions ofthe ODNs, that were reacted in Tris/HC1 buffer, to exchange with the triethylammonium cations of the mobile phase.
Oxidation of the ODN IIa/ODN tlb duplex structure. The chromatographic analysis separated the two strands of the duplex, ODN IIa eluted at 53.4 rain, ODN Ilb at 63.4 mix. The oxidized strands eluted before their corresponding non-modified strand at retention times ranging from 46.9 to 48.7 rain for ODN IIa and from 55.5 to 57.9 min for ODN Ilb. The two ofigenucleotides showed a similar conversion (50 to 70% previously to structure 221 but may rather correspond to structure 3 (see Section 4). This lesion will be now named the spiro G+34 lesion. Under the same HPLC peak, a signal at m/z 848.1 indicating a product with a loss of-133 amu from the initial ODN IIa, may correspond to an abasic site generated by the release of a guanine (or a guaninine oxidation product) by hydrolysis of the glycosidic bond.
The major HPLC peak at 47.7 rain ( Figure 2B) contains four products of ODN lla oxidation (i) the dehydro-guanidinohydantohin (1)  One remark, in ODN IIb, we observed two oxidation products eluting at different Rt (56.1 and 57.0), both corresponding to a loss of molecular mass of-39 amu, attributed to an imidazolone-containing oligonucleotide ( Table I). The fact that ODN lib cames only one guanine suggests that the two products are not due to two different positionning of the same lesion within the sequence. One hypothesis might be that these two oxidation products might correspond to t-or I-dlz nucleoside units within the oligonucleotide. :7 Oxidation of the ODN I duplex structure. Identical gu,'mine oxidation products were found on the second duplex ODN lla/ODN llb compared to the self-complementa 5"-CAGCTG (ODN I) duplex previously studied. 2 The results of the HPLC/ESI-MS analysis of the oxidation of ODN are reported on Table II under the new experimental conditions used in this work. The major lesion consisted of a product with a molecular mass increase of +4 amu, dehydro-guanidinohydantoin (1) compared to the initial mass of guanine, another product with an increase of mass of + 34 amu (3) and a product corresponding to the formation of imidazolone residue (Iz). Under the new diluted conditions used in tiffs work, the G+135 lesion was also observed as two separated peaks on ODN I." a) AM" mass of the'modified ODN minus the mass 'of the iltial ODN.
b) based on the area ofthe HPLC peak (UV detection 260 nm) In conclusion, in a reproducible manner, the oxidation of guanines within ds and ss oligonucleotides by Mn-TMPyP/KHSOs, at neutral pH and at 0 C, leads to dehydro-guanidinohydantoin derivative 1 as the main product associated to a lower amount of imidazolone lesion Iz in all the sequences tested. Modified ODNs carrying one spirt G+34 lesion are only observed in the case of the oxidation of ds ODNs. A fourth type ofDNA damage (unknown structure) appears as a lesion with a G+ 135 amu increase of mass compared to the mass of the initial ODN. This lesion was observed on single-and double-stranded oxidized DNAs. It must be noted that this lesion was not detected by us before in studies performed under different oxidation conditions. 21 2-Competition between KHSOs and H20 as trapping reagent for a common guanine cation intermediate. On the basis of previous SO labeling studies on the mechanism of the oxydation of guanine 20 21 on nucleosides and oligonucleotides by MnTMPyP/KHSOs, we proposed that an intermediate guanine cation at C5 of G was the initial event of guanine oxidation. This intermediate would undergo two competitive reaction pathways: it could be trapped by the nucleophylic attack of a molecule of water or a HSOanion. We observed that Iz did not show any tSO incorporation from labeled water on the nucleoside model. 2 This implied an exclusive nucleophilic attack of HSOat C5 of G and can be explained by the peroxide being a better nucleophile than water. KHSO5 does not exchange oxygen atoms with water. On the opposite, within ds ODNs, Iz that had incorporated one tSO, from labeled water, was the major product. We proposed that the attack ofthe HSOs anion on the guanine cation, in a double-stranded ODNs, was reduced 21 by the fact that the negatively charged peroxide has a rstricted access to the polyanionic DNA. We also know, concerning the products formed during the reaction, that Iz does not exchange its oxygen atom with solvent 2 and that compound G+34 exchanges one of its two oxygen atoms with water. 21 Table I and Table III.
Two sets of analyses were performed. Table I shows the results of labeling studies under standard conditions of mass analyses. Only the major m/z peak of the isotopic distribution was recorded. The quantitative relative abundance of the different peaks of the isotopic distributions is shown on Table III. This evaluated the competition between KHSO5 and H20 as trapping agents of the guanine cation. A higher resolution for the mass spectrometry record (0.042 amu steps) was used to obtain a clear isotope pattern for each product. In order to increase the mass detector resolution, we focused the collection ofthe data on two oxidized products, Iz and 1 containing oligonucleotides.  The results of Table I  The results of the determination of the competition of 6Ofl80 incorporation at differem KHSO5 concentrations are reported in Table III. ODNs carrying modification 1 appeared as a mixture of isomer incorporating zero to 2 atoms of 80 depending on the reaction conditions. The O incorporation on I was increased on decreasing the KHSO concentration from 500 to 100txM in both ss and ds ODN tested.
Moreover, at the same KHSO concentration, the incorporation of O from water was lower on ss DNA compared to ds DNA.
These results on the origin of the oxygen atom incorporated into i confimaed the competition between HO and HSO for the nucleophilic attack on the guanine. Decreasing the concentration of KHSO.
induced the nucleophilic attack of water to be favored (increase of O content in the products of reaction).
The ss DNA allowed a higher incorporation of 60 from KHSO due to a better accessibility of the guanines to the solvent compared to ds DNA. It is possible to observe 2 atoms of 180 as well as 2 atoms of r60 incorporated into 1. Lesion I can thus be formed from the attack of water or HSO. at C5 and C8 of G.
Moreover, these results gave a strong indication against the conversion of I into Iz. Under the conditions used in this work, Iz never showed *O incorporation. If Iz came from I it should reflect the isotopic pattern of its precursor. This is particularly clear at 100 laM KHSO, where 100% Iz appeared as 60-labeled whereas 50% 1 was doubly sO-labeled (Table III). Variation of the buffer of the reaction. The 100% trapping of the guanine cation at C5 by KHSO5 in the case ofthe Iz formation was ptzzling. This data was not in accordance with the results reported in a previous paper. : Since the present work corresponds to a reaction performed in Tris/HC1 buffer whereas in previous work labeling experiments were done in ammonium acetate buffer, we compared the incorporation of 180 into Iz in different buffers. The results are summarized in Table IV. The competition between KHSO and HO in the access to DNA was affected by the nature of the buffer. Except in Tris/HC1 buffer, Iz always showed an O incorporation ranging from 60 to 80%. The 100% 60-incorporation in Iz in Tris/HC1 buffer reported in Table III, was thus an exception. The amount of 1O incorporation in I was also lower in TrL.HCI buffer with respect to the other buffers.   deoxyribose or intrastrand 2-deoxajribose. 5 refers to previous C5 of G. The oxidized DNA strand containing one I lesion, in the case of the oxidation of the selfcomplementary 5'-CAGCTG (ODN I) 2 was purified by HPLC. Just after the collect and lyophilization procedure the HPLC/ESI-MS analysis showed the almost complete degradation of I into two degradation products, the major one (= 90%) corresponded to a loss of mas of-19 amu compared to the mass of guanine and the minor one (, 10%) to a loss of-91 amu compared to guanine (data not shown). These two products were previously observed after a heating step (90 C for 15 min) of the reaction medium of the oxidation of the same ds ODN I (200 mM ammonium acetate buffer pH 6.5, NacI 100 mM) and were proposed to be 3amino-4-carbonyl-5-[(2'-deoxy-[3-D-erythro-pentafuranosyl)amino]-2-oxoacetic acid (oxaluric acid) derivative or Oxa for the major one, and urea derivative for the minor one (Scheme 2). The conversion into these two products was complete after a heating step at 90 C for 15 min in the reaction buffer, in this case the minor presence of an abasic site was also detected. The presence of imidazolone (Iz) or its degradation product oxazolone (Z) 12'3 were never detected when the isolated oligonucleotide carrying I lesion was subjected to lyophilisation or heating. An oligonucleotide strand modified with an Iz was stable under the above described conditions. A simple explanation for the formation of these two product is given in Scheme 2 1 can lose a guanidinium moiety to give parabanic acid derivative by the hydrolysis of the imine finction. It is further hydrolyzed to give the Oxa and urea derivatives. Parabanic acid was not detected in this work.
The formation of oxaluric acid from oxidized guanidinohydantoin was also reported recently in the case of the oxidation of 8-oxo-G with singlet oxygen.  In summary it is clear that imidazolone did not arise from I as previously proposed. Compound 1 is rather the precursor of oxaluric acid derivative (Oxa) upon heating. Thus the oxidized guanidinohydantoin (1) is an independent product of guanine oxidation obtained by electron transfer. 4-The G+34 lesion corresponds to structure 3. To monitor the evolution of the G+34 lesion, the moditied oligonucleotide was collected from a preParative HPLC separation of an oxidation reaction performed on the self-complementary duplex 5'-CAGCTG (ODN I). The product was not very stable under the purification conditions. The analysis of the collected sample, recovered after collect and lyophilization, showed the presence of two new more polar products of degradation at Rt 51.6 and 53.1 min besides the peak of the spiro G+34 lesion (Rt 55.1) (Table V, entry 1). The HPLC/ESI-MS analysis showed a [M-2H] 2 signal at m/z 897.8, corresponding to a loss of 28 amu with respect to G+34, and thus corresponding to an increase of 6 arnu compared to guanine, for the product at Rt 51.6 min. The product at Rt 53.1 rain showed the same mass as the spiro G+34 (rw'z 912.0). These two products are referred to as a G+6 and a "new G+34" lesions. Heating the collected sample in water (pH 5) at 90 C for 1 h promoted the complete conversion of the spiro G+34 into the "new G+34", while the G+6 remained stable under these conditions (Table V, entry. 2). On the other hand, heating the collected sample at basic pH (pH 8.8) induced the complete conversion of the spiro G+34 into the G+6 product. The transformation was complete within 15 rnin (Table V, entry 4). The "new G+34" and G+6 lesion were stable for 1 h under these conditions (Table  V, entry 5). Therefore, complete conversion of the the spiro G+34 lesion into these two degradation products, the "new G+34" and the G+6 lesions, was achieved by heating at 90 C but was pH sensitive. The spiro G+34 was transformed into a G+6 product at pH 8.8, within 15 rain whereas it transformed into a lesion of identical mass, the "new G+34", within 1 h upon heating at 90 C in water (pH 5). Heating at 90 C at pH 7 for 1 h gave an intermediate pattern of degradation ofthe two degradation products (Table V,  entry 3). The two degradation products were stable under the different 90 C heating conditions. Mn'I3,1Pyt/KttS05 at least Three Independent Reaction Pathwm. " The so-called G+34 lesion, observed during guanine oxidation by Mn-TMPyP/K SO.. corresponds probably to structure 3 instead of structure 2 initially proposed. If the G+34 lesion is the spiro compound 3, then of course it cannot be a precursor of Iz, but is fomed via an independent reaction pathway.

CONCLUSION
The cationic metalloporphyrin, Mn-TMPyP associated to KHSO5 was used as a model reagent for the oxidation of guanine by a two-electron process. The oxidation of double-stranded oligonucleotides by Mn-TMPyP/KHSO led to three main guanine oxidation products, namely, oxidized guanidinohydantoin (1), the proposed spiro compound (3) derivative and imidazolone lesion (Iz). It was shown that these three compounds are formed by independent routes. They arise from a common precursor, namely a guanine cation generated by two electron abstracted from the guanine base. This cation exhibits two electrophilic centers (C5 and C8 carbons of guanine) that may be quenched by various nucleophiles. Depending on what nucleophile (H:O or HSOs) react at the two electrophilic carbons of guanine, the products are differem. Furthermore, the primary oxidation products, oxidized guanidinohydantoin (1), the proposed spiro compound (3) derivative and imidazolone lesion (Iz), are susceptible to further degradation. This illustrates the high complexity of the analysis of oxidative DNA damage products.