Selective abstraction of 2H from C-1' of the C residue in AGC.ICT by the radical center at C-2 of activated neocarzinostatin chromophore: structure of the drug/DNA complex responsible for bistranded lesion formation.

Glutathione-activated neocarzinostatin chromophore (NCS-Chrom) generates bistranded lesions at AGC.GCT sequences in DNA, consisting of an abasic site at the C residue and a strand break at the T residue on the complementary strand, due to hydrogen atom abstraction from C-1' and C-5', respectively. Earlier work showed that 2H from C-5' of T was selectively abstracted by the radical center at C-6 of activated NCS-Chrom, supporting a proposed model of the active-drug/DNA complex. However, since under the conditions used breaks at the T exceeded their inclusion in bistranded lesions, it was not clear what fraction of the hydrogen transfer represented bistranded lesions. Since virtually all abasic sites at the C are part of a bistranded lesions, hydrogen transfer from C-1' of C into the drug should reflect only the bistranded reaction. Accordingly, a self-complementary oligodeoxynucleotide 5'-GCAGCICTGC-3' was synthesized in which the C contained 2H at the C-1' position. In order to eliminate an 2H isotope effect on the transfer and to increase the extent of the bistranded reaction, an I residue was substituted for the G opposite the C residue. Sequencing gel electrophoretic analysis revealed that under one-hit kinetics, 37% of the damage reaction was associated with abasic site (alkali-labile break) formation at the C residue and 48% with direct strand breaks at the T residue. Thus, 74% of the damage involved a bistranded lesion. 1H NMR spectroscopic analysis of the reacted chromophore showed that 2H had been selectively transferred into the C-2 position to the extent of approximately 22%.(ABSTRACT TRUNCATED AT 250 WORDS)     

Free radical mechanisms in neocarzinostatin-induced DNA damage

The molecular mechanisms by which the antitumor protein antibiotic, neocarzinostatin, interacts with DNA and causes DNA sugar damage is discussed. Physical binding of the nonprotein chromophore of neocarzinostatin to DNA, involving an intercalative process and dependent on the microheterogeneity of DNA structure, is followed by thiol activation of the drug to a probable radical species. The latter attacks the deoxyribose, especially at thymidylate residues, by abstracting a hydrogen atom from C-5' to generate a carbon-centered radical on the DNA. This nascent form of DNA damage either reacts with dioxygen to form a peroxyl radical derivative, which eventuates in a strand break with a nucleoside 5'-aldehyde at the 5'-end or reacts with the bound drug to form a novel drug-deoxyribose covalent adduct. Nitroaromatic radiation sensitizers can substitute for dioxygen, but the DNA damage products are different. Similarities between the various biological effects of neocarzinostatin and ionizing radiation are reviewed.     

C-1' hydrogen abstraction of deoxyribose in DNA strand scission by dynemicin A.

Dynemicin A, which is a hybrid antitumor antibiotic containing anthraquinone and enediyne cores, abstracts the C-1' hydrogen of DNA deoxyribose and then the damaged DNA leads to strand breaks with the formation of 5'- and 3'-phosphate termini. The lesions of C-4' hydrogen also occur at 3' side of G.C base pairs (i. e., 5'-CT and 5'-GA), leading to 5'-phosphate and 3'-phosphoglycolate termini or 4'-hydroxylated abasic sites. The C-1' hydrogen abstraction by dynemicin A is distinct from the preferential C-5' hydrogen abstraction of calicheamicin and neocarzinostatin.     

Detection of neocarzinostatin chromophore-deoxyribose adducts as exonuclease-resistant sites in defined-sequence DNA

A 5'-end-labeled DNA restriction fragment was treated with the nonprotein chromophore of neocarzinostatin under anoxia in the presence of dithiothreitol, conditions known to maximize formation of chromophore-deoxyribose adducts. Under conditions where unmodified DNA was digested to completion, chromophore-treated DNA was highly resistant to digestion by exonuclease III plus the 3'----5' exonucleolytic activity of T4 DNA polymerase and partially resistant to digestion by exonuclease III plus snake venom exonuclease. The electrophoretic mobilities of the products of exonucleolytic digestion suggested that (i) digestion by exonuclease III or T4 polymerase terminated one nucleotide before the nucleotide containing the adduct, (ii) the remaining nucleotide directly adjacent to the adduct (3' side) could be removed by snake venom phosphodiesterase, but at a slow rate, (iii) the covalently linked chromophore decreased the electrophoretic mobilities of the digestion products by the equivalent of approximately three nucleotides, and (iv) adducts formed under anaerobic conditions occurred at the same nucleotide positions as the strand breaks formed under aerobic conditions (primarily at T and, to a lesser extent, A residues). The close similarity in sequence specificity of adducts and strand breaks suggests that a common form of nascent DNA damage may be a precursor to both lesions. A chromophore-induced free radical on C-5' of deoxyribose, subject to competitive fixation by addition reactions with either oxygen or chromophore, is the most likely candidate for such a precursor. The base specificity of adduct formation does not reflect the reported base specificity of neocarzinostatin-induced mutagenesis, suggesting that lesions other than adducts may be responsible for at least some neocarzinostatin-induced mutations, particularly those occurring at G X C base pairs.     

Poly(deoxyadenylic-deoxythymidylic acid) damage by radiolytically activated neocarzinostatin

The anaerobic reaction of poly(deoxyadenylic-deoxythymidylic acid) with neocarzinostatin activated by the carboxyl radical CO2-, an electron donor generated from gamma-ray radiolysis of nitrous oxide saturated formate buffer, has been characterized. DNA damage includes base release and strand breaks. Few strand breaks are formed prior to alkaline treatment; they bear 3'-phosphoryl termini. In contrast, most (66%) of the base release occurs spontaneously. DNA damage is highly (95%) specific for thymidine sites. Neither DNA-drug covalent adduct nor nucleoside 5'-aldehyde, which are major products in the DNA-nicking reaction initiated by mercaptans and oxygen, is formed in this reaction. Data are presented to show that the CO2(-)-activated neocarzinostatin intermediate is a short-lived free radical able to abstract hydrogen atoms from the C-1' and C-5' positions of deoxyribose. Attack occurs mostly (68%) at the C-1' position, producing a lesion whose properties are consistent with those of (oxidized) apyrimidinic sites.     

Atypical abasic sites generated by neocarzinostatin at sequence-specific cytidylate residues in oligodeoxynucleotides

Neocarzinostatin chromophore produces alkali-labile, abasic sites at cytidylate residues in AGC sequences in oligonucleotides in their duplex form. Glutathione is the preferred thiol activator of the drug in the formation of these lesions. The phosphodiester linkages on each side of the abasic site are intact, but when treated with alkali, breaks are formed with phosphate moieties at each end. Similar properties are exhibited by the abasic lesions produced at the purine residue to which the C in AGC is base-paired on the complementary strand. The abasic sites at C residues differ from those produced by acid-induced depurination in the much greater lability of the phosphodiester linkages on both sides of the deoxyribose, in the inability of NaBH4 to prevent alkali-induced cleavage, and in the relative resistance to apurinic/apyrimidinic endonucleases. The importance of DNA microstructure in determining attack site specificity in abasic site formation at C residues is shown not only by the requirement for the sequence AGC but also by the findings that substitution of G by I 5' to the C decreases the attack at C, whereas placement of an I opposite the C markedly enhances the reaction. Quantitation of the abstraction of 3H into the drug from C residues in AGC specifically labeled in the deoxyribose at C-5' or C-1',2' suggests that, in contrast to the attack at C-5' in the induction of direct strand breaks at T residues, abasic site formation at C residues may involve attack at C-1'. Each type of lesion may exist on the complementary strands of the same DNA molecule, forming a double-stranded lesion.     

Lysine 2,3-aminomutase. Support for a mechanism of hydrogen transfer involving S-adenosylmethionine

The conversion of L-lysine to L-beta-lysine is catalyzed by lysine 2,3-aminomutase. The reaction involves the interchange of the 2-amino group of lysine with a hydrogen at carbon 3. As such the reaction is formally analogous to adenosylcobalamin-dependent rearrangements. However, the enzyme does not contain and is not activated by this coenzyme. Instead it contains iron and pyridoxal phosphate and is activated by S-adenosylmethionine. Earlier experiments implicated adenosyl-C-5' of S-adenosylmethionine in the hydrogen transfer mechanism, apparently in a role similar or analogous to that of adenosyl moiety of adenosylcobalamin in the B12-dependent rearrangements. The question of whether both hydrogens or only one hydrogen at adenosyl-C-5' participate in the hydrogen-transfer process has been addressed by carrying out the lysine 2,3-aminomutase reaction with S-[5'-3H] adenosylmethionine in the presence of 10 times its molar concentration of enzyme. Under these conditions all of the tritium appeared in lysine and beta-lysine, showing that C-5'-hydrogens participate. To determine whether hydrogen transfer is compulsorily intermolecular and intramolecular, various molar ratios of [3,3-2H2]lysine and unlabeled lysine were submitted to the action of lysine 2,3-aminomutase under conditions in which 10-15% conversion to beta-lysine occurred. Mass spectral analysis of the beta-lysine for monodeutero and dideutero species showed conclusively that hydrogen transfer is both intramolecular and intermolecular. The results quantitatively support our postulate that activation of the enzyme involves a transformation of S-adenosylmethionine into a form that promotes the generation of an adenosyl-5' free radical, which abstracts hydrogen from lysine to form 5'-deoxyadenosine as an intermediate.     

The effect of reductant levels on the formation of damage lesions derived from a 2-deoxyribose radical in ssDNA

Thiols play a major role in the outcome of oxidative damage to DNA when it is initiated through cellular exposure to ionizing radiation. DNA radicals formed under aerobic conditions are converted to peroxyl radicals through trapping by oxygen at a diffusion-controlled rate. As a primary source of cellular reductant, thiols are responsible for the conversion of these DNA-derived peroxyl radicals to their corresponding hydrogen peroxides and subsequent strand breaks. Through the use of modified nucleotides, which act as precursors to nucleic acid radicals, we have investigated the effect of varying amounts of the cellular thiol glutathione (GSH) on the distribution of damage products produced from a 2-deoxyribose radical in DNA: the C3'-thymidinyl radical. The C3'-thymidinyl radical results from the abstraction of a hydrogen atom from the C3'-position of DNA oligomers at a thymidine residue, and is known to deliver several DNA damage lesions including the 3'-phosphoglycolaldehyde, 3'-phosphoglycolate and a 5'-aldehyde. Here we show that the level of GSH present has an impact on the level of production of these C3'-thymidinyl radical derived damage products.     

Mechanistic studies on the base-catalyzed transformation of neocarzinostatin chromophore: roles of bulged DNA

Nucleic acid bulges have been implicated in a number of biological processes and are specific cleavage targets for the enediyne antitumor antibiotic neocarzinostatin chromophore (NCS-chrom) in a base-catalyzed, radical-mediated reaction. Studies designed to elucidate the detailed mechanism of the base-catalyzed activation of NCS-chrom and to evaluate the roles of bulged DNA in its activation are described. They show that nucleobases in the DNA bulge are not required to form an effective bulge pocket but enhance the binding of the wedge-shaped activated drug molecule. Analysis of solvent deuterium isotope effects on NCS-chrom degradation and DNA cleavage efficiency experiments suggests that the spirolactone biradical 6 is a relatively stable species and that intramolecular quenching of the C2 radical of 6 to form the biologically active cyclospirolactone radical 7a occurs first (pathway a in Scheme 2), leaving the C6 radical to abstract the hydrogen atom from the DNA deoxyribose and to form the cyclospirolactone 8. Binding of the activated drug at the bulge site is required, but not sufficient, for efficient 8 formation, whereas cleavage of bulged DNA is not essential. Efficient generation of 8, but inefficient DNA damage generation, comes mainly from the likely high off-rate of 7a binding. The finding that thymidine 5'-carboxylic acid-ended oligonucleotide fragment can be formed in the reaction suggests that the process of DNA cleavage is rather slow and that sequential oxidations of the target 5'-carbon are possible. Study of the effect of solvent (methanol) concentration on NCS-chrom degradation indicates that bulged DNA acts to assist the intramolecular quenching of the radical at C2 by C8' ' of the naphthoate moiety by excluding solvent from the binding pocket, thus preventing the formation of spirolactones 9, and by blocking radical polymerization. Because in the absence or near absence of solvent methanol 8 formation does not reach even 10% that formed in the presence of bulged DNA, it is possible that the DNA bulge also induces a conformational change in the drug to promote the intramolecular reaction.     

Product analyses in DNA strand scission by antitumor antibiotic elsamicin A.

Elsamicin A is an antitumor antibiotic with fascinating chemical structure and a good candidate for pharmaceutical development. Molecular mechanism of DNA backbone cleavage mediated by Fe(II)-elsamicin A has been examined. Product analysis using DNA sequencing gels and HPLC reveals the production of damaged DNA fragments bearing 3'-/5'-phosphate and 3'-phosphoglycolate termini associated with formation of free base. In addition, hydrazine-trapping experiments indicate that C-4' hydroxylated abasic sites are formed concomitant with DNA degradation by Fe(II)-elsamicin A. The results lead to the conclusion that the hydroxyl radical formed in Fe(II)-elsamicin A plus dithiothreitol system oxidizes the deoxyribose moiety via hydrogen abstraction predominantly at the C-4' carbon of the deoxyribose backbone and ultimately produces strand breakage of DNA.     

The interaction of adriamycin and adriamycin analogues with nucleic acids in the B and A conformations.

The reinforced intercalative binding to DNA typical of adriamycin and daunomycin can still occur if there is epimerisation at C4' or if the O-methyl group is lost or if the 9-substituents are deleted or if the 4'-hydroxyl group is lost. In the latter two cases however, there is a reduction in affinity for the DNA, supporting the suggested role of the 9-hydroxyl and 4'-hydroxyl groups in secondary stabilization of the complex. Epimerisation at C-1' or at C-3' alters but does not abolish the intercalative mode of binding to DNA whereas epimerisation at C-7 precludes intercalation of the chromophore into the helix of DNA. In contrast to the interaction with the B-form found in DNA, the parent drugs do not intercalate into nucleic acids possessing the A-conformation and none of the above-mentioned structural changes will allow intercalation into A-form nucleic acids.     

Oxygen transfer from the nitro group of a nitroaromatic radiosensitizer to a DNA sugar damage product

Mechanisms based on one-electron oxidation appear incomplete in explaining cellular radiosensitization by nitroaromatic compounds such as misonidazole. Evidence is presented for a novel mechanism that may be involved in enhancing DNA strand breakage due to a variety of agents, including ionizing radiation, that generate carbon-centered radicals on DNA deoxyribose. Under anaerobic conditions the carbon-centered radical generated selectively at C-5' of deoxyribose of thymidylate residues in DNA by the antitumor antibiotic neocarzinostatin reacts with misonidazole to produce a DNA damage product in the form of 3'-(formyl phosphate)-ended DNA. In an 18O-transfer experiment we find that the carbonyl oxygen of the activated formyl moiety (trapped as formyl-Tris) is derived from the nitro group oxygen of misonidazole. This result strongly supports a mechanism in which a nitroxide radical adduct, formed by the addition of misonidazole to the radical at C-5' of deoxyribose, cleaves between the N and O so as to form an oxy radical precursor of the formyl moiety and a two-electron reduction species of misonidazole.     

Generation of superoxide free radical by neocarzinostatin and its possible role in DNA damage

Spectroscopic analysis of the reduction of both nitro blue tetrazolium and ferricytochrome c induced by neocarzinostatin shows that superoxide free radical is produced during the spontaneous degradation of the antibiotic. The amount of superoxide free radical produced from neocarzinostatin is not affected by the presence of thiol, although earlier work has shown that DNA damage is stimulated at least 1000-fold by thiol. Transition metals are not involved in this reaction. Although superoxide dismutase inhibits the reduction of nitro blue tetrazolium and cytochrome c induced by neocarzinostatin, neither it nor catalase interferes with the action of neocarzinostatin on DNA, whether or not drug has been activated by thiol. The pH profiles for spontaneous base release and alkali-labile base release (a measure of nucleoside 5'-aldehyde formation at a strand break) do not correspond with that for the generation of superoxide free radical from neocarzinostatin. The same holds for supercoiled DNA cutting by neocarzinostatin chromophore in the absence of a thiol, which is an acid-favored reaction. These results indicate that the generation of superoxide free radical by the drug does not correlate with DNA damage activity, whether or not thiol is present. Furthermore, the failure of hydroxyl free-radical scavengers to inhibit drug-induced single-strand breaks in supercoiled DNA in the absence of thiol also indicates that a diffusible hydroxyl free radical is most probably not involved in this reaction.     

The reaction of the substrate analog 2-ketoglutarate with adenosylcobalamin-dependent glutamate mutase

Glutamate mutase is one of several adenosylcobalamin-dependent enzymes that catalyze unusual rearrangements that proceed through a mechanism involving free radical intermediates. The enzyme exhibits remarkable specificity, and so far no molecules other than L-glutamate and L-threo-3-methylaspartate have been found to be substrates. Here we describe the reaction of glutamate mutase with the substrate analog, 2-ketoglutarate. Binding of 2-ketoglutarate (or its hydrate) to the holoenzyme elicits a change in the UV-visible spectrum consistent with the formation of cob(II)alamin on the enzyme. 2-ketoglutarate undergoes rapid exchange of tritium between the 5'-position of the coenzyme and C-4 of 2-ketoglutarate, consistent with the formation of a 2-ketoglutaryl radical analogous to that formed with glutamate. Under aerobic conditions this leads to the slow inactivation of the enzyme, presumably through reaction of free radical species with oxygen. Despite the formation of a substrate-like radical, no rearrangement of 2-ketoglutarate to 3-methyloxalacetate could be detected. The results indicate that formation of the C-4 radical of 2-ketoglutarate is a facile process but that it does not undergo further reactions, suggesting that this may be a useful substrate analog with which to investigate the mechanism of coenzyme homolysis.     

The photochemistry of 8-bromo-2'-deoxyadenosine. A direct entry to cyclopurine lesions.

The UV photolysis of 8-bromo-2'-deoxyadenosine has been investigated in different solvents and in the presence of additives like halide anions. Photolytic cleavage of the C-Br bond leads to formation of the C8 radical. In methanol, subsequent hydrogen abstraction from the solvent is the main radical reaction; however, in water or acetonitrile intramolecular hydrogen abstraction from the sugar moiety, to give the C5' radical, is the major path. This C5' radical undergoes a cyclization reaction on the adenine and gives the aminyl radical. A rate constant of 1.8 x 10(5) s(-1) has been measured by laser flash photolysis in CH(3)CN for this unimolecular process. Product studies from steady-state photolysis in acetonitrile have shown the conversion of 8-bromo-2'-deoxyadenosine to 5',8-cyclo-2'-deoxyadenosine in 65% yield and in a diastereoisomeric ratio (5' R):(5' S)= 1.7. Evidence supporting that the equilibrium Br*+ Br(-)[right left harpoons] Br(2)*(-) plays an important role in this synthetically useful radical cascade is obtained by regulating the relative concentrations of the two reactive oxidizing species.     

Incorporation of 18O2 into thymidine 5'-aldehyde in neocarzinostatin chromophore-damaged DNA

Strand scission of DNA by the chromophore of neocarzinostatin converts the 5'-hydroxyl of deoxyribose to a 5'-aldehyde. The origin of the aldehydic oxygen has now been elucidated by mass spectrometry. DNA-associated thymidine 5'-aldehyde produced by treatment of DNA with neocarzinostatin chromophore in 2H218O/16O2 or in 2H216O/18O2 was reduced, liberated by nuclease treatment, permethylated, and analyzed by gas chromatography-mass spectrometry. The data clearly show that molecular oxygen is the only source of the 5'-aldehydic oxygen. The addition of molecular oxygen at C-5', possibly via a reactive form of neocarzinostatin chromophore, must be involved; a carbonium ion intermediate at C-5' is ruled out.     

The influence of protonation or alkylation of the phosphate group on the e.s.r. spectra and on the rate of phosphate elimination from 2-methoxyethyl phosphate 2-yl radicals.

The e.s.r. spectra of 1-yl, 2-yl and 3'-yl methoxyethyl phosphate radicals derived from CH3OCH2CH2-OPO3H2 by hydrogen abstraction have been measured in aqueous solutions and the hyperfine constants determined. The coupling constants vary strongly with protonation or alkylation of the phosphate group. The 2-yl radicals eliminate phosphate. The rate-constants for the elimination (ke) have been estimated by e.s.r. measurements and by product studies as a function of pH using 60Co gamma-radiolysis. The ke values vary from approximately 0.3 s(-1) for the CH3OCHCH2OPO3--radical and approximately 10(3) s-1 for CH3OCHCH2OPO3H-, to approximately 3 X 10(6) S-1 for CH3OCHCH2OPO3H2. Alkylation of the phosphate group increases the elimination rate-constant to a similar extent as protonation. The results support a recent mechanism which described the OH-radical-induced single-strand breaks of DNA in aqueous solution starting from the C-4' radical of the sugar moiety. It is further concluded the C-4' radical of DNA eliminates the 3'-phosphate group faster than the 5'-phosphate group.     

Mechanisms of DNA damage by chromium(V) carcinogens.

Reactions of bis(2-ethyl-2-hydroxy-butanato)oxochromate(V) with pUC19 DNA, single-stranded calf thymus DNA (ss-ctDNA), a synthetic oligonucleotide, 5'-GATCTATGGACTTACTTCAAGGCCGGGTAATGCTA-3' (35mer), deoxyguanosine and guanine were carried out in Bis-Tris buffer at pH 7.0. The plasmid DNA was only nicked, whereas the single-stranded DNA suffered extensive damage due to oxidation of the ribose moiety. The primary oxidation product was characterized as 5-methylene-2-furanone. Although all four bases (A, C, G and T) were released during the oxidation process, the concentration of guanine exceeds the other three. Orthophosphate and 3'-phosphates were also detected in this reaction. Likewise, the synthetic oliogomer exhibits cleavage at all bases with a higher frequecncy at G sites. This increased cleavage at G sites was more apparent after treating the primary oxidation products with piperidine, which may indicate base oxidation as well. DNA oxidation is shown to proceed through a Cr(V)-DNA intermediate in which chromium(V) is coordinated through the phosphodiester moiety. Two alternative mechanisms for DNA oxidation by oxochromate(V) are proposed to account for formation of 5-methylene-2-furanone, based on hydrogen abstraction or hydride transfer from the C1' site of the ribose followed by hydration and two successive beta-eliminations. It appears that phosphate coordination is a prerequisite for DNA oxidation, since no reactions between chromium(V) and deoxyguanosine or guanine were observed. Two other additional pathways, hydrogen abstraction from C4' and guanine base oxidation, are also discussed.     

Biosynthesis of 9-beta-D-arabinofuranosyladenine: hydrogen exchange at C-2' and oxygen exchange at C-3' of adenosine

The data presented here describe new findings related to the bioconversion of adenosine to 9-beta-D-arabinofuranosyladenine (ara-A) by Streptomyces antibioticus by in vivo investigations and with a partially purified enzyme. First, in double label in vivo experiments with [2'-18O]- and [U-14C]adenosine, the 18O:14C ratio of the ara-A isolated does not change appreciably, indicating a stereospecific inversion of the C-2' hydroxyl of adenosine to ara-A with retention of the 18O at C-2'. In experiments with [3'-18O]- and [U-14C]-adenosine, [U-14C]ara-A was isolated; however, the 18O at C-3' is below detection. The adenosine isolated from the RNA from both double label experiments has essentially the same ratio of 18O:14C. Second, an enzyme has been isolated and partially purified from extracts of S. antibioticus that catalyzes the conversion of adenosine, but not AMP, ADP, ATP, inosine, guanosine, or D-ribose, to ara-A. In a single label enzyme-catalyzed experiment with [U-14C]adenosine, there was a 9.9% conversion to [U-14C]ara-A; with [2'-3H]-adenosine, there was a 8.9% release of the C-2' tritium from [2'-3H]adenosine which was recovered as 3H2O. Third, the release of 3H as 3H2O from [2'-3H]adenosine was confirmed by incubations of the enzyme with 3H2O and adenosine. Ninety percent of the tritium incorporated into the D-arabinose of the isolated ara-A was in C-2 and 8% was in C-3. The enzyme-catalyzed conversion of adenosine to ara-A occurs without added cofactors, displays saturation kinetics, a pH optimum of 6.8, a Km of 8 X 10(-4) M, and an inhibition by heavy metal cations. The enzyme also catalyzes the stereospecific inversion of the C-2' hydroxyl of the nucleoside antibiotic, tubercidin to form 7-beta-D-arabinofuranosyl-4-aminopyrrolo[2,3-d]pyrimidine. The nucleoside antibiotic, sangivamycin, in which the C-5 hydrogen is replaced with a carboxamide group, is not a substrate. On the basis of the single and double label experiments in vivo and the in vitro enzyme-catalyzed experiments, two mechanisms involving either a 3'-ketonucleoside intermediate or a radical cation are proposed to explain the observed data.