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.     

Molecular models of neocarzinostatin damage of DNA: analysis of sequence dependence in 5''GAGCG:5''CGCTC.

Model building and molecular mechanics and dynamics calculations have been performed on a number of complexes of the post-activated form of the neocarzinostatin chromophore (NCS) with the B-DNA oligomer 5'GAGCG:5'CGCTC. Stable structures with the naphthoic acid moiety intercalated at all base pairs can be constructed. The observed bistranded lesions consisting of an abasic site at the Cyt residue in AGC and a direct break at the Thy residue on the complementary strand can be explained by assuming that NCS in the (R,R) form intercalates between the Ade2-Thy9/Gua3-Cyt8 base step with its 'diradical' core oriented towards the 3'-end of the (+) strand. Sites at C5', C4' and C1' in the minor groove are within a short enough distance from the two radical centers on NCS to permit hydrogen atom abstraction and the formation of the bistranded lesions. Strand cleavage at Thy9 may occur as a single lesion if NCS is intercalated into the Gua3-Cyt8/Cyt4-Gua7 base step with its active core towards the 3'-end of the (-) strand. The results are analyzed, and the utility and limitations of this type of model building are discussed.     

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.     

Mismatch-induced switch of neocarzinostatin attack sites in the DNA minor groove.

Based on the finding that the wobble G.T mismatch 5' to the C of AGC.GCT results in switching of the attack chemistry by neocarzinostatin chromophore (NCS-Chrom) on the deoxyribose moiety of C from C-1' to C-4' [Kappen, L. S. & Goldberg, I. H. (1992) Proc. Natl. Acad. Sci. U.S.A. 89, 6706-6710], a series of mismatches has been explored for their effect on the chemistry of damage at the T of AGT.ACT in oligodeoxynucleotides, a site at which 4'-chemistry ordinarily occurs. Placement of a G.T mispair 5' to the T results in a marked increase in 4'-chemistry, as measured by the formation of breaks with 3'-phosphoglycolate ends and abasic sites due to 4'-hydroxylation. Strikingly, 4'-chemistry is induced at the T on the complementary strand, a site ordinarily restricted to 5'-chemistry. Substitution of dioxygen by the radiation sensitizer misonidazole exerts a pronounced effect on the partitioning of the 4'-chemistry in favor of the 3'-phosphoglycolate product. Both stable T.G and unstable T.C mismatches at the attack site itself are associated with marked inhibition of damage at this site. Whereas placement of the relatively stable G.A mismatch on the 5'-side of the T residue (AGT) results in substantial inhibition of damage at the T without shifting of chemistry, the same mismatch at the 3'-side of the attack site decreases damage only slightly but is associated with the appearance of significant 1'-chemistry. By contrast, no shift in chemistry is found with bleomycin, which attacks at C-4'.(ABSTRACT TRUNCATED AT 250 WORDS)     

Activation of neocarzinostatin chromophore and formation of nascent DNA damage do not require molecular oxygen.

Thiol-activated neocarzinostatin chromophore abstracts tritium from the 5', but not from the 1' or 2' positions of deoxyribose in DNA and incorporates it into a stable, non-exchangeable form. The abstracted tritium remains covalently associated with the chromophore or its degradation product after treatment with acid or alkali, respectively. Drug activation and the consequent hydrogen abstraction reaction, presumably generating a carbon-centered radical at C-5', do not require molecular oxygen but have a dose-dependent relation with thiol. Under aerobic conditions, where base release and DNA strand breaks with nucleoside 5'-aldehyde at the 5'-ends are produced, hydrogen abstraction from C-5' parallels these parameters of DNA damage. It is possible to formulate a reaction scheme in which the carbon- centered radical at C-5' is an intermediate in the formation of the various DNA damage products found under both aerobic and anaerobic conditions.     

Bistranded oxidized purine damage clusters: induced in DNA by long-wavelength ultraviolet (290-400 nm) radiation?

Bistranded clustered DNA damages involving oxidized bases, abasic sites, and strand breaks are produced by ionizing radiation and radiomimetic drugs, but it was not known whether they can be formed by other agents, e.g., nonionizing radiation. UV radiation produces clusters of cyclobutyl pyrimidine dimers, photoproducts that occur individually in high yield. Since long-wavelength UV (290-400 nm) radiation induces oxidized bases, abasic sites, and strand breaks at low yields, we tested whether it also produces clusters containing these lesions. We exposed supercoiled pUC18 DNA to UV radiation with wavelengths of >290 nm (UVB plus UVA radiation), and assessed the induction of bistranded clustered oxidized purine and abasic clusters, as recognized by Escherichia coli Fpg protein and E. coli Nfo protein (endonuclease IV), respectively, as well as double-strand breaks. These three classes of bistranded clusters were detected, albeit at very low yields (37 Fpg-OxyPurine clusters Gbp(-1) kJ(-1) m(2), 8.1 double-strand breaks Gbp(-1) kJ(-1) m(2), and 3.4 Nfo-abasic clusters Gbp(-1) kJ(-1) m(2)). Thus, these bistranded OxyPurine clusters, abasic clusters, and double-strand breaks are not uniquely induced by ionizing radiation and radiomimetic drugs, but their level of production by UVB and UVA radiation is negligible compared to the levels of frequent photoproducts such as pyrimidine dimers.     

Induction of double-strand breaks by S1 nuclease, mung bean nuclease and nuclease P1 in DNA containing abasic sites and nicks.

Defined DNA substrates containing discrete abasic sites or paired abasic sites set 1, 3, 5 and 7 bases apart on opposite strands were constructed to examine the reactivity of S1, mung bean and P1 nucleases towards abasic sites. None of the enzymes acted on the substrate containing discrete abasic sites. Under conditions where little or no non-specific DNA degradation was observed, all three nucleases were able to generate double-strand breaks when the bistranded abasic sites were 1 and 3 base pairs apart. However, when the abasic sites were further apart, the enzymes again failed to cleave the DNA. These results indicate that single abasic sites do not cause sufficient denaturation of the DNA to allow incision by these single-strand specific endonucleases. The reactivity of these enzymes was also investigated on DNA substrates that were nicked by DNasel or more site-specifically by endonuclease III incision at the discrete abasic sites. The three nucleases readily induced a strand break opposite such nicks.     

Sequence-specific, strand-selective, and directional binding of neocarzinostatin chromophore to oligodeoxyribonucleotides

The sequence-specific interaction of neocarzinostatin chromophore (NCS-C) has been evaluated with a series of synthetic oligodeoxyribonucleotides of defined sequences containing the most preferred nucleotide cleavage site, T or A, or both. NCS-C preferentially cleaves T or A residues in the sequence GN1N2, where N2 is T or A. Greater cleavage occurs on the strand enriched with G residues, provided that they are adjacent to other G residues, but not at N1. These results are compatible with a model for drug binding in which the naphthoate moiety of NCS-C preferentially intercalates at GN1. This is accompanied by electrostatic binding interaction provided by the positively charged amino sugar moiety so as to place the reactive bicyclo[7.3.0]dodecadienediyne epoxide moiety in an appropriate orientation in the minor groove enabling, upon thiol activation, attack at C-5' of T or A. At certain sequences, such as GCT.AGC, a similar binding mode is also able to generate a basic lesions at the C residue on the opposite strand, forming a bistranded lesion. Although the reactions with glutathione generally show the same strand selectivity and sequence specificity as those with dithiothreitol, the former is usually more efficient than the latter.     

Probing DNA single strands for single-base bulges with neocarzinostatin chromophore.

Neocarzinostatin chromophore (NCS-Chrom) induces strong cleavage at a single site (C3) in the single-stranded and 5' (32)P-end-labeled 13-mer GCCAGATTTGAGC in a reaction dependent on a thiol. By contrast, in the duplex form of the same 13-mer, strand cleavage occurs only at the T and A residues, and C3 is not cleaved. To determine the minimal structural requirement(s) for C3 cleavage in the single-stranded oligomer, several deletions and mutations were made in the 13-mer. A 10-mer (GCCAGAGAGC) derived from the 13-mer by deletion of the three T residues was also cleaved exclusively at C3 by NCS-Chrom, generating fragments having 5' phosphate ends. That the cleavage at C3 is initiated by abstraction of its 5' hydrogen is confirmed in experiments using 3' (32)P-end-labeled 10-mer. The competent 13-mer and 10-mer were assigned hairpin structures with a stem loop and a single bulged out A base, placing C3 across from and 3' to the bulge. Removal of the bulged A base from the 13-mer and the 10-mer resulted in complete loss of cutting activity, proving that it is the essential determinant in competent substrates. Studies of thiol post-activated NCS-Chrom binding to the DNA oligomers show that the drug binds to the bulge-containing 13-mer (K(d) = 0.78 microM) and the 10-mer (K(d) = 1.11 microM), much more strongly than to the 12-mer (K(d) = 20 microM) and the 9-mer (K(d) = 41 microM), lacking the single-base bulge. A mutually induced-fit between NCS-Chrom and the oligomer resulting in optimal stabilization of the drug-DNA complex is proposed to account for the site-specific cleavage at C3. These studies establish the usefulness of NCS-Chrom as a probe for single-base bulges in DNA.     

NMR solution structures of bistranded abasic site lesions in DNA

Ionizing radiation produces clustered lesions in DNA. Since the orientation of bistranded lesions affects their recognition by DNA repair enzymes, clustered damages are more difficult to process and thus more toxic than single oxidative lesions. In order to understand the structural determinants that lead to differential recognition, we used NMR spectroscopy and restrained molecular dynamics to solve the structure of two DNA duplexes, each containing two stable abasic site analogues positioned on opposite strands of the duplex and staggered in the 3' (-1 duplex, (AP) 2-1 duplex) or 5' (+1 duplex, (AP) 2+1 duplex) direction. Cross-peak connectivities observed in the nonexchangeable NOESY spectra indicate compression of the helix at the lesion site of the duplexes, resulting in the formation of two abasic bulges. The exchangeable proton spectra show the AP site partner nucleotides forming interstrand hydrogen bonds that are characteristic of a Watson-Crick G.C base pairs, confirming the extra helical nature of the AP residues. Restrained molecular dynamics simulations generate a set of converging structures in full agreement with the spectroscopic data. In the (AP) 2-1 duplex, the extra helical abasic site residues reside in the minor groove of the helix, while they appear in the major groove in the (AP) 2+1 duplex. These structural differences are consistent with the differential recognition of bistranded abasic site lesions by human AP endonuclease.     

Neocarzinostatin-induced hydrogen atom abstraction from C-4' and C-5' of the T residue at a d(GT) step in oligonucleotides: shuttling between deoxyribose attack sites based on isotope selection effects.

The thiol-activated neocarzinostatin chromophore cleaves duplex oligonucleotides containing the sequence-TGTTTGA-, producing 3'-phosphoglycolate and 3'-phosphate fragments at T, indicating the involvement of 4'- as well as 5'-chemistry at this residue. Substitution of deuterium for hydrogen at the C-4' position of the affected T leads to a kinetic isotope effect (kH/kD) of 4.0 on the formation of the glycolate-ended product, whereas deuterium at C-5' of the same T reveals kH/kD of 1.6 in the formation of the phosphate-ended product. The proportion of the products representing 4'- and 5'-chemistry can be shifted on the basis of isotope selection effects. A second product resulting from 4'-chemistry, the abasic site associated with 4'-hydroxylation, has been identified as an alkali-labile site, and as a pyridazine derivative formed after cleavage by hydrazine. A comparable isotope effect on its production (kH/kD = 3.7) relative to that of 3'-phosphoglycolate production is consistent with a common intermediate, a putative 4'-peroxy radical, in their formation. The formation of both products of 4'-chemistry is oxygen-dependent, and the internal partitioning between them (3'-phosphate or 3'-phosphoglycolate) is influenced by thiols. Moreover, the nitroaromatic radiation sensitizer misonidazole can substitute for dioxygen, yielding 3'-phosphoglycolate and alkali-labile 3'-phosphate ends, indicative of 4'-chemistry. In addition to the internal partitioning of 4'-chemistry, thiols also affect the overall extent of cleavage (4' plus 5') and the relative partitioning between both sites of attack (4' or 5').     

Quantification of DNA strand breaks and abasic sites by oxime derivatization and accelerator mass spectrometry: application to gamma-radiation and peroxynitrite

We report a highly sensitive method to quantify abasic sites and deoxyribose oxidation products arising in damaged DNA. The method exploits the reaction of aldehyde- and ketone-containing deoxyribose oxidation products and abasic sites with [(14)C]methoxyamine to form stable oxime derivatives, as originally described by Talpaert-Borle and Liuzzi [Reaction of apurinic/apyrimidinic sites with [(14)C]methoxyamine. A method for the quantitative assay of AP sites in DNA, Biochim. Biophys. Acta 740 (1983) 410-416]. The sensitivity of the method was dramatically improved by the application of accelerator mass spectrometry to quantify the (14)C, with a limit of detection of 1 lesion in 10(6) nucleotides in 1 microg of DNA. The method was validated using DNA containing a defined quantity of abasic sites, with a >0.95 correlation between the quantities of abasic sites and those of methoxyamine labels. The original applications of this and similar oxyamine derivatization methods have assumed that abasic sites are the only aldehyde-containing DNA damage products. However, deoxyribose oxidation produces strand breaks and abasic sites containing a variety of degradation products with aldehyde and ketone moieties. To assess the utility of methoxyamine labeling for quantifying strand breaks and abasic sites, the method was applied to plasmid DNA treated with gamma-radiation and peroxynitrite. For gamma-radiation, there was a 0.99 correlation between the quantity of methoxyamine labels and the quantity of strand breaks and abasic sites determined by a plasmid nicking assay; the abasic sites comprised less than 10% of the radiation-induced DNA damage. Studies with peroxynitrite demonstrate that the method, in conjunction with DNA repair enzymes that remove damaged bases to produce aldehydic sugar residues or abasic sites, is also applicable to quantifying nucleobase lesions in addition to strand break products. Compared to other abasic site quantification techniques, the modified method offers the advantage of providing a straightforward and direct measurement of aldehyde- and ketone-containing strand breaks and abasic sites, with the potential for direct labeling in cells prior to DNA isolation.     

In vitro repair of synthetic ionizing radiation-induced multiply damaged DNA sites.

When ionizing radiation traverses a DNA molecule, a combination of two or more base damages, sites of base loss or single strand breaks can be produced within 1-4 nm on opposite DNA strands, forming a multiply damaged site (MDS). In this study, we reconstituted the base excision repair system to examine the processing of a simple MDS containing the base damage, 8-oxoguanine (8-oxoG), or an abasic (AP) site, situated in close opposition to a single strand break, and asked if a double strand break could be formed. The single strand break, a nucleotide gap containing 3' and 5' phosphate groups, was positioned one, three or six nucleotides 5' or 3' to the damage in the complementary DNA strand. Escherichia coli formamidopyrimidine DNA glycosylase (Fpg), which recognizes both 8-oxoG and AP sites, was able to cleave the 8-oxoG or AP site-containing strand when the strand break was positioned three or six nucleotides away 5' or 3' on the opposing strand. When the strand break was positioned one nucleotide away, the target lesion was a poor substrate for Fpg. Binding studies using a reduced AP (rAP) site in the strand opposite the gap, indicated that Fpg binding was greatly inhibited when the gap was one nucleotide 5' or 3' to the rAP site.To complete the repair of the MDS containing 8-oxoG opposite a single strand break, endonuclease IV DNA polymerase I and Escherichia coli DNA ligase are required to remove 3' phosphate termini, insert the "missing" nucleotide, and ligate the nicks, respectively. In the absence of Fpg, repair of the single strand break by endonuclease IV, DNA polymerase I and DNA ligase occurred and was not greatly affected by the 8-oxoG on the opposite strand. However, the DNA strand containing the single strand break was not ligated if Fpg was present and removed the opposing 8-oxoG. Examination of the complete repair reaction products from this reaction following electrophoresis through a non-denaturing gel, indicated that a double strand break was produced. Repair of the single strand break did occur in the presence of Fpg if the gap was one nucleotide away. Hence, in the in vitro reconstituted system, repair of the MDS did not occur prior to cleavage of the 8-oxoG by Fpg if the opposing single strand break was situated three or six nucleotides away, converting these otherwise repairable lesions into a potentially lethal double strand break.     

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.     

Condensation of DNA—A putative obstruction for repair process in abasic clustered DNA damage

Clustered DNA damages are defined as two or more closely located DNA damage lesions that may be present within a few helical turns of the DNA double strand. These damages are potential signatures of ionizing radiation and are often found to be repair resistant. Types of damaged lesions frequently found inside clustered DNA damage sites include oxidized bases, abasic sites, nucleotide dimers, strand breaks or their complex combinations. In this study, we used a bistranded two-lesion abasic cluster DNA damage model to access the repair process of DNA in condensate form.Oligomer DNA duplexes (47 bp) were designed to have two deoxyuridine in the middle of the sequences, three bases apart in opposite strands. The deoxyuridine residues were converted into abasic sites by treatment with UDG enzyme creating an abasic clustered damage site in a precise position in each of the single strand of the DNA duplex. This oligomer duplex having compatible cohesive ends was ligated to pUC19 plasmid, linearized with HindIII restriction endonuclease. The plasmid–oligomer conjugate was transformed into condensates by treating them with spermidine. The efficiency of strand cleavage action of ApeI enzyme on the abasic sites was determined by denaturing PAGE after timed incubation of the oligomer duplex and the oligomer–plasmid conjugate in presence and absence of spermidine. The efficiency of double strand breaks was determined similarly by native PAGE. Quantitative gel analysis revealed that rate of abasic site cleavage is reduced in the DNA condensates as compared to the oligomer DNA duplex or the linear ligated oligomer–plasmid conjugates. Generation of double strand break is significantly reduced also, suggesting that their creation is not proportionate to the number of abasic sites cleaved in the condensate model. All these suggest that the ApeI enzyme have difficulty to access the abasic sites located deep into the condensates leading to repair refractivity of the damages. In addition, we found that presence of a polyamine such as spermidine has no notable effect in the incision activity of ApeI enzyme in linear oligomer DNA duplexes in our experimental concentration.     

Spectrum of complex DNA damages depends on the incident radiation

Ionizing radiation induces bistranded clustered damages--two or more abasic sites, oxidized bases and strand breaks on opposite DNA strands within a few helical turns. Since clusters are refractory to repair and are potential sources of double-strand breaks (DSBs), they are potentially lethal and mutagenic. Although induction of single-strand breaks (SSBs) and isolated lesions has been studied extensively, little is known about the factors affecting induction of clusters other than DSBs. To determine whether the type of incident radiation could affect the yields or spectra of specific clusters, we irradiated genomic T7 DNA, a simple 40-kbp linear, blunt-ended molecule, with ion beams [iron (970 MeV/nucleon), carbon (293 MeV/nucleon), titanium (980 MeV/nucleon), silicon (586 MeV/nucleon), protons (1 GeV/nucleon)] or 100 kVp X rays and then quantified DSBs, Fpg-oxypurine clusters and Nfo-abasic clusters using gel electrophoresis, electronic imaging and number average length analysis. The yields (damages/Mbp Gy(-1)) of all damages decreased with increasing linear energy transfer (LET) of the radiation. The relative frequencies of DSBs compared to abasic and oxybase clusters were higher for the charged particles-including the high-energy, low-LET protons-than for the ionizing photons.     

Modulation of neocarzinostatin-mediated DNA double strand damage by activating thiol: deuterium isotope effects.

The neocarzinostatin chromophore causes double-strand damage at AGC sequences on DNA by concomitant 1'-oxidation at C and 5'-oxidation at the T on the complementary strand. The extent of this damage is dependent upon the structure of the thiol used for activation. Deuterium isotope effects suggest that this dependence on thiol structure may be due to internal quenching of one radical site of the activated chromophore by the hydrogen atoms of the thiol sidechain. The 12-mer d[GCAAGCGCTTGC] is treated with the neocarzinostatin chromophore and either sodium thioglycolate or [2-2H2]-thioglycolate, and the distribution of strand breaks is determined by gel electrophoresis. Two isotope effects are noted: an overall sequence-independent effect in which deuterated thioglycolate increases total strand damage by a factor of 2, and a sequence-specific effect by which deuteration increases the proportion of alkali-sensitive strand damage at C6 by an additional factor of 1.5. Methyl thioglycolate shows essentially identical behavior to that of thioglycolate anion, ruling out electrostatic effects as major contributors to the effect of thiol structure on the mode of DNA damage observed. A model for NCSC action consistent with these results is 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.