HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Henle, E. S. Articles by Linn, S. Search for Related Content PubMed PubMed Citation Articles by Henle, E. S. Articles by Linn, S. Social Bookmarking                      What's this?

J Biol Chem, Vol. 274, Issue 2, 962-971, January 8, 1999

Sequence-specific DNA Cleavage by Fe²⁺-mediated Fenton Reactions Has Possible Biological Implications^*

Ernst S. Henle, Zhengxu Han, Ning Tang^§, Priyamvada Rai, Yongzhang Luo^¶, and Stuart Linn

From the Division of Biochemistry and Molecular Biology, University of California, Berkeley, California 94720-3202

	ABSTRACT

Top Abstract Introduction Procedures Results Discussion References

Preferential cleavage sites have been determined for Fe²⁺/H₂O₂-mediated oxidations of DNA. In 50 mM H₂O₂, preferential cleavages occurred at the nucleoside 5' to each of the dG moieties in the sequence RGGG, a sequence found in a majority of telomere repeats. Within a plasmid containing a (TTAGGG)₈₁ human telomere insert, 7-fold more strand breakage occurred in the restriction fragment with the insert than in a similar-sized control fragment. This result implies that telomeric DNA could protect coding DNA from oxidative damage and might also link oxidative damage and iron load to telomere shortening and aging. In micromolar H₂O₂, preferential cleavage occurred at the thymidine within the sequence RTGR, a sequence frequently found to be required in promoters for normal responses of many procaryotic and eucaryotic genes to iron or oxygen stress. Computer modeling of the interaction of Fe²⁺ with RTGR in B-DNA suggests that due to steric hindrance with the thymine methyl, Fe²⁺ associates in a specific manner with the thymine flipped out from the base stack so as to allow an octahedrally-oriented coordination of the Fe²⁺ with the three purine N⁷ residues. Fe²⁺-dependent changes in NMR spectra of duplex oligonucleotides containing ATGA versus those containing AUGA or A^5mCGA were consistent with this model.

	INTRODUCTION

Top Abstract Introduction Procedures Results Discussion References

The cytotoxicity of hydrogen peroxide is mediated by highly reactive oxidants generated by the reaction of reduced transition metals with H₂O₂ via Fenton reactions (1). DNA is a major target in H₂O₂-mediated cell killing through such reactions (2, 3), but the oxidizing species involved are probably not freely diffusible hydroxyl radicals (^·OH), but possibly a localized ^·OH and/or related iron-oxo species (4-8) whose properties are apparently governed by the chelation state of the iron upon which they are generated (9). Because DNA can chelate Fe²⁺ in several ways, DNA damage in vivo in prokaryotes (2) and eukaryotes (10, 11), and in vitro (2, 9) exhibit peculiar dose responses to H₂O₂ concentration. The rates of killing or DNA damage induction are maximal below about 1 mM, then drop off and become independent of H₂O₂ concentration between approximately 5 and 50 mM. Based upon these responses, we have defined two types of oxidants, termed types I and II. Type I oxidants are scavenged by H₂O₂ to form less reactive species, exemplified by the reaction of ^·OH with H₂O₂. Hence, these oxidants appear to damage DNA only at lower H₂O₂ concentrations; they are also scavenged by alcohols. The constant rate of damage observed with H₂O₂ concentrations between 5 and 50 mM was defined to be due to type II oxidants. In this case DNA damage might be induced by the reaction of H₂O₂ with Fe²⁺ atoms that are intimately bound to DNA such that nascent oxidants react with the DNA before encountering exogenous scavengers such as H₂O₂ or ethanol (9). Alternatively, one could propose that type II oxidants do not react with H₂O₂ or ethanol. In either case, these oxidants would be formed on Fe²⁺ ions that are associated differently within the DNA helix than those which give rise to type I oxidants.

Because of the high reactivity of ^·OH and related oxidants, damage to DNA in vivo or in the presence of scavengers in vitro is thought to occur predominantly with such radicals when they are generated by reaction of H₂O₂ with metal ions bound to the DNA (12-14). The binding of metals to preferred sequences within DNA and the relationship of such associations to DNA damage by the Fenton reaction have been the subject of recent investigations (13, 15-17), and it has been proposed that transition metals in the presence of H₂O₂ might cause sequence-specific damage to DNA (12, 18). However, whereas iron associated with DNA has been repeatedly implicated in cytotoxic DNA damage (1, 3, 12, 14, 19, 20), sequence preference for damage has been reported only with copper/ascorbate (17) copper/NADH (21), with Fe²⁺ bound to chelators or larger molecules (22), or with bleomycin (23). In fact, Henner et al. (24) reported that the distribution of strand breaks induced by excess Fe²⁺ (160 nM DNA-nucleotide, 1 mM Fe²⁺, no added H₂O₂) was indistinguishable from the distribution upon treatment with -radiation and that strand breakage was distributed equally (±20%) among the four nucleotides. However, the DNA had been stored in distilled H₂O and it is likely that this DNA was not in a double-stranded, B-form.

In contrast, our laboratory found that nucleoside damage following exposure of DNA to Fe²⁺ and H₂O₂ was not equally distributed among the nucleosides (25). Since we have proposed that DNA damage in 0.5 mM H₂O₂ was caused by Fe²⁺ ions that associate differently with DNA than those which cause damage at 50 mM H₂O₂, we have now investigated whether strand breaks vary in sequence location at the two H₂O₂ concentrations. In this study, we demonstrate that preferential cleavage sites do in fact differ for type I (0.5 mM H₂O₂) and type II (50 mM H₂O₂) oxidants, and that these sites can be grouped into small consensus sequences. These consensus sequences, which appear to reflect specific modes of iron binding by DNA, might be involved in regulatory processes of the cell.

	EXPERIMENTAL PROCEDURES

Top Abstract Introduction Procedures Results Discussion References

Materials-- H₂O was deionized and then distilled. H₂O₂ was a 30% solution from Fisher Scientific; FeSO₄·7H₂O and NaCl (99.999%) were from Aldrich; ethanol (100%) was from Quantum Chemical Co. The highly purified NaCl was essential in order to avoid the accumulation of transition metals on DNA substrates. H₂O₂ and Fe²⁺ concentrations were measured as described previously (26). Reagents used for the Maxam-Gilbert sequencing were from Sigma and were freshly prepared according to Maxam and Gilbert (1977) (27). Restriction enzymes were from New England Biolabs (Beverly, MA); polynucleotide kinase was from Promega (Madison, WI).

CGCGATATGACACTAG, CGCGATAUGACACTAG, CTAGTGTCATATCGCG, CGCGATA^5mCGACACTAG, and CTAGTGTCGTATCGCG were obtained from Operon Technologies (Alameda, CA) and were individually purified on a C18 HPLC¹ column (10 mm × 25 cm, Econosphere; Alltech, Deerfield, IL) using a 5-15% acetonitrile gradient in 50 mM triethylammonium acetate, pH 7.3, and then repeatedly desiccated in the presence of ammonia to remove triethylamine. Each 16'-mer was then loaded onto the same column through which 1 M NaCl was pumped for 30 min followed by 30 min of H₂O and then finally the sodium form DNA was eluted with 50% methanol (HPLC grade, Fisher). Duplexes were formed in 130 mM NaCl by the method of continuous fractions (28) and verified to be free of excess single-stranded oligonucleotide by NMR.

A plasmid containing human telomeric DNA, (TTAGGG)₈₁, was a gift from Dr. Carol Greider, Cold Spring Harbor Laboratory. Plasmid pBluescript SK was from Stratagene, and plasmid pBSRep81 was constructed by inserting the telomere into the pBluescript SK. Escherichia coli XL1-Blue was used for plasmid maintenance and propagation. The plasmid DNA was purified by a Qiaprep Spin plasmid miniprep kit (Qiagen, Chatsworth, CA). The DNA was dialyzed at 4 °C against 1 mM EDTA, 50 mM NaCl for 24 h and then against 50 mM NaCl for 24 h.

Isolation of 144- and 191-bp PM2 DNA Fragment Substrates-- Bacteriophage PM2 DNA was isolated as described by Kuhnlein et al. (29), dialyzed against 0.1 mM sodium EDTA, 10 mM Tris-HCl, pH 7.5, for 2 days, and then extensively dialyzed against 130 mM NaCl. It was restricted with HindIII, and then the digest was separated on a 1.5% agarose gel into seven fragments of roughly 100, 250, 425, 450, 1000, 2100, and 5300 bp. A band containing the unresolved 425- and 450-bp fragments was cut out and the DNA electro-eluted into a dialysis bag with 0.089 M Tris borate, 2 mM EDTA, pH 7.8 (TBE). The eluate was extracted with 2-butanol (Aldrich), the volume was reduced to 0.5 ml, and then the DNA was precipitated three times with ethanol. The DNA pellet was dissolved in TBE and digested with HaeIII, and the digest was resolved on a 2.2% agarose gel.

With HaeIII, the 425-bp fragment generates 144- and 191-bp fragments from the ends and a 90-bp fragment from the middle. (The 450-bp HindIII fragment generates 2 bands of about 68- and 382-bp.) To prepare 5'-³³P-labeled 144- and 191-bp DNA fragments, the mixture of the 425- and 450-bp HindIII fragments was labeled with polynucleotide kinase and [-³³P]ATP before cleavage with HaeIII. The end-labeled 144- and 191-bp fragments were then electrophoretically separated on an 8% polyacrylamide gel. The gel slices were individually electro-eluted into TBE buffer, and the eluted DNA was extracted with phenol and chloroform and then precipitated with ethanol and re-dissolved in 65 mM NaCl. Each fragment was then dialyzed for 1 day each against 2 mM sodium EDTA, 5 mM Tris-HCl, pH 7.5, and 0.2 mM EDTA, 5 mM Tris-HCl, pH 7.5, and finally extensively against 65 mM NaCl (99.999%). The DNA was free of contamination by transition metals as detected by the absence of nicking in the presence of H₂O₂. To label the 3'-ends of the fragments, the HindIII restriction fragments were treated with Klenow DNA polymerase in the presence of 50 µM each dATP and dGTP, and 40 µCi (0.5 µM) of [-³³P]dCTP (2000 Ci/µmol, NEN Life Science Products), and then the DNA was processed as described for the 5'-labeled fragments. The initial sequencing of the 144- and 191-bp fragments was performed on both strands by the dideoxy procedure according to the protocol supplied by United States Biochemicals.

DNA Damage Protocol for Determination of Strand Break Locations-- A solution containing 2.5-8 µM end-labeled restriction fragment in 65 mM NaCl, 20 nM to 2 µM FeSO₄, 65 mM NaCl, and 10 mM ethanol as indicated, was brought to a final concentration of either 0.5 or 50 mM H₂O₂, and incubated for 25 min at room temperature. A large excess of 0.3 M sodium acetate, pH 6.5, was then added, and the DNA was precipitated with ethanol. The samples were centrifuged for 45 min at 23,000 × g, and then the pellets were suspended and reprecipitated twice with ethanol and finally dissolved in 7 µl of sequencing buffer (27). After heating at 90 °C for 2 min, the resuspended damaged DNA was loaded onto a pre-warmed 40-cm-long 8% polyacrylamide gel in 7 M urea and electrophoresed at 60 watts for 3-4 h. Autoradiography was performed by exposure of the dried gel to Kodak XAR-5 film for 3-10 days at 70 °C. Quantitation was accomplished by scanning the autoradiogram and determining the area beneath a peak using ImageQuant version 1.1 (Molecular Dynamics). For sequence markers, Maxam-Gilbert sequencing reactions of undamaged substrate DNAs were loaded onto the same gel. For detecting double strand DNA breaks after Fenton reactions, the reacted DNA was analyzed on 8% polyacrylamide gels.

Protocol for Nicking of Plasmids-- The indicated amount of FeSO₄ was added to 0.1 pmol of DNA in 130 mM NaCl and incubated at room temperature for the times indicated. Ethanol (100%) was then added where indicated to a final concentration of 10 mM (type I) or 100 mM (type II), and the oxidation was initiated by adding H₂O₂ to a final concentration of 50 µM (type I) or 50 mM (type II). The total volume was 10 µl. After 30 min at room temperature, EDTA was added to a final concentration of 50 µM and then forms I, II, and III of the plasmid DNA were separated by electrophoresis on 1% agarose gels. The gels were stained with ethidium bromide and also imaged, and the percentage of form I DNA quantified using an AlphaImager (Alpha Innotech, San Leandro, CA). Ethidium bromide stains form I one third less efficiently than forms II or III, and calculations were adjusted accordingly. The number of nicks per molecule was calculated from the amount of form I remaining, assuming a Poisson distribution.

Labeling and Restriction of Nicked Plasmid DNA-- Ten units of T4 polynucleotide kinase and 2 pmol of [-³³P]ATP were incubated with 0.1 pmol of plasmid DNA in 50 mM Tris-HCl (pH 7.6) and 10 mM MgCl₂. After 30 min at 37 °C, the kinase was inactivated by heating to 75 °C for 30 min and the labeled plasmid DNA was digested with either 10 units of HinfI in 50 mM NaCl, 10 mM Tris-HCl (pH 7.9), 10 mM MgCl₂, 1 mM dithiothreitol or with 10 units of NciI in 50 mM potassium acetate, 10 mM magnesium acetate, 1 mM dithiothreitol (pH 7.9) at 37 °C for 60 min. The digests were separated by electrophoresis on 2% agarose gels. The gels were stained, imaged, and quantified as above, and subsequently dried, and the bands were imaged and quantified using a PhosphorImager and ImageQuant version 1.1.

	RESULTS

Top Abstract Introduction Procedures Results Discussion References

Strand Break Locations with Fe²⁺ and 50 mM H₂O₂ (Type II Oxidants)-- In order to study sequence preferences by types I and II radicals during iron-mediated Fenton reactions, we examined the location of cleavages after treating 144- and 191-base pair PM2 DNA restriction fragments with Fe²⁺ and either 0.5 or 50 mM H₂O₂ (Fig. 1).

View larger version (50K): [in this window] [in a new window]   Fig. 1.   Sequences and preferential cleavage sites of the 144- and 191-bp PM2 DNA restriction fragments. Preferential cleavage sites that occur in the presence of 0.5 mM H2O2 () and 50 mM H2O2 () are shown. Asterisks indicate the location of the 3'- or 5'-33P labels used for the study. Lowercase lettering indicates nucleotides in regions of the gels that were usually not sufficiently resolved to unambiguously determine sites of nicking. Underscored lettering indicates the region that is shown in Fig. 2.

As predicted, cleavage sites differed for the two concentrations. In addition, the cleavages were predominantly within a few short specific sequences. When 50 mM H₂O₂ was present, preferential cleavage occurred at the nucleosides 5' to dG moieties in the consensus sequence, RGGG (Figs. 1 and 2, Table I), at some dC moieties in the sequence RCR, and at one dA moiety within the sequence YAAGA. (Bold underscored nucleotides indicate the sites of cleavage.) In the RGGG consensus sequence, the extent of cleavage was generally heaviest the more 5' the nucleotide resided in the sequence and the 3' dGMP was not preferentially cleaved. Three dC moieties were identified as preferential cleavage sites, two of which were in the consensus sequence, RCR (Table I, Figs. 1 and 2). Since 38 other dC residues in RCR were not preferentially cleaved under these conditions, this sequence by itself obviously does not constitute a preferential cleavage site. The extent of strand breakage in the presence of 50 mM H₂O₂ was not diminished by the presence of 100 mM ethanol, as would be expected for damage by a type II oxidant. When the DNA was exposed to Fe²⁺ and 50 mM H₂O₂ and then analyzed on polyacrylamide gels lacking urea, no evidence for the occurrence of double strand breaks was found (data not shown).

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Examples of preferential cleavages. A, a sample autoradiogram that contains examples of preferential cleavages at (R)GGG, RTGR, YAAGT, and RCR within the 5'-labeled 144-bp fragment. Reactions for lanes 1-3 contained 0.5 mM H2O2 and 0.1, 0.5, and 2 µM Fe2+, respectively. Reactions for lanes 4 and 5 each contained 50 mM H2O2 and 0.1 µM Fe2+; that in lane 5 also contained 100 mM ethanol. Based on previous studies (2, 3, 9, 26), we estimate that the maximum amount of strand breakage due to the Fe2+ and H2O2 is less than 2%. The lane marked C contains a control sample with Fe2+ and H2O2 omitted. Lanes marked C/T or A/G are Maxam-Gilbert sequence reactions of the same fragment for pyrimidines and purines, respectively. Numbers in parentheses following "5'" or "3'" are the bp number in the fragments as shown in Fig. 1. The sequence illustrated here is underlined in Fig. 1. This is one representative experiment; gels for each fragment with 3'- or 5'-end-labeled were obtained at least twice under all conditions. B, the autoradiogram shown in A was digitized with a ScanMaker IIHR (Microtek, Redondo Beach, CA). The lanes were converted to profiles by ImageQuant version 1.1 (Molecular Dynamics, Sunnyvale, CA). The peaks (bands) of a given lane were normalized by the amount of loaded DNA as determined by the band corresponding to the added fragment originating from the 450-bp HindIII fragment, which was generated by labeling of that fragment along with the 425-bp precursor to the 144-bp fragment, and digestion with HaeIII (see "Experimental Procedures"). The integrated peak areas were corrected by subtracting out the normalized amount found for the control sample. Each bar represents the corrected peak area corresponding to the indicated nucleotide. The corrected data for lanes 2 and 4 are shown as examples.

Strand Break Locations with Fe²⁺ and 0.5 mM H₂O₂ (Type I Oxidants)-- With 0.5 mM H₂O₂, small amounts of cleavage occurred at RGGG and RCR sites, as would be expected if type II radicals were to form to a limited extent in 0.5 mM H₂O₂ (Table I, Figs. 1 and 2). However, preferential cutting in 0.5 mM H₂O₂ occurred mainly within the consensus sequences RTGR, TATTY, CTTR, and YAAGT. Also, as expected for cleavages by type I radicals, 10 mM ethanol inhibited all of these preferential cleavages (except those at RGGG) (Table I, Figs. 1 and 2). No double-strand breaks were observed. It is to be noted that controls (Fig. 2) include both the omission of H₂O₂ and, perhaps more significantly, the utilization of concentrations of Fe²⁺ that were too low to give rise to detectable breaks, even in the presence of H₂O₂. Five dC residues were cleaved in the RCR consensus sequence; however, 35 other dC residues in RCR were not preferentially cleaved under these conditions. Again, this sequence by itself obviously does not constitute a preferential cleavage site for type I oxidants.

The sequence RTGR was generally the most strongly cleaved. It appears 9 times in the two fragments and in all but one case, preferential cleavage was evident at the dT moiety (Table I; Fig. 1). (In the exception, 191-bp fragment, nucleotides 49-53, the RTGR is flanked by two cleavable TATTT sequences and opposite a cleaved ACG sequence.) The RTGR consensus sequence appears to be quite stringent, since only 2 of the 40 sequences in the two substrate DNAs that differ by one nucleotide (RT(A/Y)R, YTGR, or RTGY) and are not part of other cleavage sequences were preferential thymidine cleavage sites (Table I, Fig. 1).

All five TATTY sequences were preferential cleavage sites (Table I, Figs. 1 and 2). Breaks were also made at the two CATTR sites (base pairs 26 and 136, 191-bp fragment). These latter sequences may be grouped as non-consensus variants of either the RTGR or the TATTY consensus sequences. Alternately, these sites may constitute a separate consensus sequence.

The seven CTTR sequences present in the two fragments constitute preferential cleavage sites (Table I, Fig. 1 and 2C) for the 3' dTMP residues. Three additional preferential cleavage sites (base pairs 25 and 91 in the 144-bp fragment and base pair 80 in the 191-bp fragment) are non-consensus variants of this sequence.

Preferential cleavage at the 3' dAMP in the YAAGT was found in four of six of these sequences, albeit with less efficiency than with the aforementioned dTMP residues (Table I, Figs. 1 and 2). Two non-consensus variants, TAAGC (base pair 50, 144-bp fragment) and TGAGC (base pair 57, 144-bp fragment), may broaden the specificity of this consensus to RYAGY.

It is noteworthy that in only two cases (191-bp fragment, bp 98; 144-bp fragment, bp 60) was cleavage noted at a nucleotide that was 3' to a dCMP residue. Such a nick would have left a sugar fragment attached to dCMP, and such residues were noted by Bertoncini and Meneghini (11), to be uniquely absent from the termini of damaged DNA resulting from the treatment of CV1-P cells with H₂O₂. In addition, in no case were preferential cleavage sites immediately 5' to a dCMP, and in only 1 of 50 sites was there a pyrimidine nucleotide both 3' and 5' to the cleavage site (within the T8 run at bp 118 of the 144-bp fragment). Finally, all preferential cleavages at a purine nucleotide were 5' to another purine nucleotide. Cleavages are clearly far from random, and purine nucleotides strongly favor the presence of adjacent preferential cleavages, whereas cytosine nucleotides strongly discourage them.

Preferential Nicking of Telomeric DNA on a Plasmid by Type II Radicals-- Preferential cleavages by Fe²⁺/H₂O₂ at specific sequences could be due to preferential association of the iron at that sequence and/or due to electronic effects that attract and "trap" radicals formed elsewhere, as described by Hall et al. (30). The almost absolute specificity for nicking at RGGG in higher H₂O₂ concentrations is potentially significant, given that this sequence is characteristic of the large majority of telomeric repeats (31). (Exceptions among the telomeres of some fungi often have repeats containing RTGR.) Therefore, in order to determine whether telomeric DNA within a plasmid would sensitize the plasmid to type II strand breakage, we subjected an equimolar mixture of pBluescript plasmid and a pBluescript plasmid (pBSRep81) that contains an insert of a human telomere, (TTAGGG)₈₁, to 50 mM H₂O₂ in the presence of Fe²⁺ and 10 mM ethanol. While the number of nicks per base pair increased with ferrous ion concentration, it remained equal between the two plasmids (Fig. 3). This equality held whether the plasmid was pre-incubated with iron for 5 or 120 min before H₂O₂ addition. Hence, while RGGG was identified as a preferential cleavage site, the telomere-containing plasmid did not sustain more nicks per base pair than the plasmid without the insert. However, at very high ferrous ion concentrations, considerably more double strand breakage was observed on the plasmid containing the telomere insert than on the plasmid without the insert (Fig. 3), suggesting that the single strand breaks may have been unequally distributed within the plasmid containing the telomere insert.

View larger version (42K): [in this window] [in a new window]   Fig. 3.   Relative nicking in 50 mM H2O2 of plasmids with or without a telomere insert. Lane 1 contains pBluescript (2.96 kilobase pairs), and lane 2 contains pBSrep81 (3.45 kilobase pairs; the pBluescript plasmid with the telomere insert). For lanes 3-9, pBluescript and pBSRep81 were mixed in equal amounts as determined by UV absorption and then 0.1 pmol of this DNA in 0.8% NaCl was incubated with FeSO4 for 5 min at the final concentrations shown. Reactions were initiated by addition of ethanol and H2O2 to final concentrations of 10 and 50 mM, respectively, and the reaction mixtures were allowed to incubate for 30 min at room temperature before being quenched with 50 µM EDTA. The ethidium bromide-stained gel was imaged and digitized. The tabulated results from lanes 4-6 are corrected for the initial nicks present as determined for lane 3. Lanes 7-9 did not have sufficient (if any) form I for single-strand break assessment but are amenable to visual estimation of the relative double strand breakage between the two plasmids, as indicated by the appearance of form III (linear) DNA.

To investigate the distribution of breaks within the telomere-containing plasmid, we exposed the plasmid to H₂O₂/Fe²⁺ and then radiolabeled the strand breaks with T4 polynucleotide kinase and [-³³P]ATP. After inactivation of the kinase, the plasmid was digested with HinfI. Among the restriction fragments was a 842-bp fragment that contained the 480-bp telomere insert and a total of 87 RGGG sequences (81 from the telomere and 6 from plasmid DNA). As a control fragment, we monitored a 1074-bp fragment that contained 13 RGGG sites. Ethidium bromide staining and quantitation confirmed complete digestion by the HinFI and quantitated the DNA loaded and separated by agarose gel electrophoresis. Autoradiography of the separated fragments allowed a determination of the nicks put into each of the restriction fragments (Fig. 4A).

View larger version (48K): [in this window] [in a new window]   Fig. 4.   Distribution of strand breaks within the plasmid containing a telomere insert. A, all reactions contained 0.1 pmol of pBSRep81 and 0.8% NaCl. Lane 1 contains undamaged DNA which was radiolabeled and then digested with HinfI. The slight PhosphorImager intensities of the bands (which are not visible in the figure) were subtracted for calculations. Lane 2 contains plasmid DNA exposed for 5 min to 300 nM FeSO4 before the addition of 10 mM ethanol and 50 mM H2O2 (final concentrations). The reaction mixture was incubated for 30 min at room temperature and then quenched with 50 µM EDTA, and then the products were labeled with polynucleotide kinase and [-33P]ATP and finally digested with HinfI. The reaction in lane 3 was the same as that in lane 2 except that 2000 nM FeSO4 was present. Lane 4 contains an undamaged HinFI digest, which was 33P-labeled. B, lane 1 contains undamaged pBSrep81, which was 33P-end-labeled and then digested with HinfI. Lane 2 contains plasmid DNA exposed for 120 min to 300 nM FeSO4 before the addition of 10 mM ethanol and 50 mM H2O2 (final concentrations). The reaction mixture was incubated for 30 min at room temperature and quenched with 50 µM EDTA, and then the products were 33P-end-labeled and finally digested with HinfI. Lane 3 contains an undamaged HinfI digest which was 33P-end-labeled. Lane 4 contains an unlabeled 1-kilobase ladder (GenSura Laboratories, Inc.). C, lane 1 contains 0.1 pmol of pBSRep81 in 0.8% NaCl exposed for 120 min to 300 nM FeSO4 before the addition of H2O2 to 50 µM. The reaction mixture was incubated for 30 min at room temperature and then quenched with 50 µM EDTA, and then the products were 33P-end-labeled and digested by HinfI. Lane 2 contains an undamaged HinfI digest, which was 33P-end-labeled.

When the plasmid was pre-incubated for 5 min at room temperature with the Fe²⁺ prior to addition of H₂O₂, the fragment containing the telomeric DNA sustained 1.6-fold the number of nicks (Fig. 4A) of the control fragment. However, when the pre-incubation period was extended to 120 min, the telomeric DNA sustained 7-fold the number of nicks (Fig. 4B) of the control fragment. Increasing the pre-incubation period to 18 h did not further increase the ratio of nicks of the fragment with the telomeric DNA to that of the control fragment (data not shown). The ratio of RGGG sites containing the telomeric DNA to the control fragment is 6.7 (87:13), which corresponds to the aforementioned 7-fold increase in nicks. When the damaged and ³³P-labeled DNA was digested by NciI, a similar ratio equal to the relative content of RGGG sites was observed, which depended upon pre-incubation of Fe²⁺ and DNA. In summary, it appears that Fe²⁺ ions are attracted to the two plasmids based upon their relative sizes. Then, in a time-dependent manner, these ions search out and bind to the RGGG sequences, which are targets for type II radicals.

These results support the observation that RGGG is the preferential cleavage site for type II oxidations and provides further evidence for the prediction that the iron-DNA associations that lead to type II oxidations are due to iron imbedded in the base stack next to purines (9). The requirement for pre-incubation of the Fe²⁺ and DNA would suggest a minimal contribution of the RGGG sequence acting as a sink of electrons (or electron holes) traversing the base stack, and favors specific Fe²⁺ binding.

Because the number of sensitive sites for type I radicals is large and includes sequences other than RGGG, the relative reactivity of the telomere-containing restriction fragment was investigated in 50 µM H₂O₂. The DNA and Fe²⁺ concentrations were kept the same as above, and the Fe²⁺ and DNA were pre-incubated for 120 min at room temperature before the addition of the 50 µM H₂O₂. The ratio of the labeling of the telomere-containing 842-bp fragment was 1.6-fold that of the control, 1074-bp fragment (Fig. 4C). This value is in agreement with the relative content of consensus sequences for type I radicals, RTGR, TATTY, CTTR, YAAGT, and RGGG, which is 101 versus 62, or a ratio of 1.63.

The Interaction of Fe²⁺ with the RTGR Consensus Sequence-- Among the preferential cleavages in micromolar H₂O₂ concentrations, cleavage at the RTGR thymidine was observed generally to be the strongest. In addition, as noted under "Discussion," RTGR appears to be a necessary promoter element for the regulation of certain genes which are responsive to oxidative or iron stress. Since the RTGR sequence has no apparent electronic distinctions, we explored ways in which Fe²⁺ might interact directly with the sequence ATGA in a duplex oligonucleotide, using the InsightII molecular modeling program (Fig. 5). Generally, the major groove is more distinctive than the minor groove for base distinction, particularly since the pyrimidines have C² keto groups in the minor groove. However, attempts to place the Fe²⁺ between the N⁷ of the 5'-A of ATGA and that of the G in the major groove showed that the methyl group of thymine sterically interfered. Given the precedent of "base flipping" by the restriction methylases and repair DNA glycosylases (32, 33), we allowed the thymine to "flip out" of the base stack, possibly having been displaced by the Fe²⁺ or more likely to have spontaneously flipped out during "breathing" so as to allow the Fe²⁺ to enter. In this energy minimization modeling (Fig. 5A), the Fe²⁺ was allowed initially to coordinate the guanine N⁷. Coordination by the guanine O⁶ could then also occur via a water bridge (34), and Fe²⁺ is also attracted to the two phosphates that flank the thymidine. Steric hindrance between the Fe²⁺ and the thymine methyl in B DNA stabilizes the thymidine in the displaced position. The 5'-adenine then repositions such that its N⁷ takes part in the iron chelation. If this chelation is restricted to approximately 2 Å, surprisingly the 3'-adenine also repositions such that its N⁷ is within chelation distance of the iron. The three purine N⁷ residues and the phosphate between the dG and the dT would thus form a chelation pocket that is suitable for an octahedrally coordinated iron (Fig. 5B). If the guanine O⁶ also coordinates the iron via a water bridge, 5 of the 6 iron coordination sites are occupied. The last site, which is accessible to solvent, is juxtaposed to the thymidine deoxyribose which is the preferential cleavage site in the presence of 50 µM H₂O₂. Finally, the opposite strand remains essentially undistorted (Fig. 5, C and D).

View larger version (0K): [in this window] [in a new window]   Fig. 5.   Modeling of the interaction of Fe2+ with the sequence ATGA in a duplex oligonucleotide. A-D, Distortion of the sodium form B-DNA duplex 8'-mer 5'-ATATGACA/3'-TATACTGT due to interactions of iron with the ATGA was simulated using the Discover module of the Insight II version 95.0 computer program (Biosym Technologies). Enhanced by steric hindrance from the thymine methyl group, the dT moiety was allowed to flip out of the helix as explained in the text. While constraining the most 5' and 3' base pairs, the N7 groups of the 5'-adenine and the guanine in the ATGA consensus sequence were constrained to approximately 2 Å from the iron. However, the angles of this simulated chelation were not constrained. A, a schematic diagram representing the simulated mechanism for an iron-RTGR interaction. Angles and distances are not to scale. Fe2+ moves along the phosphate backbone and is weakly chelated by the O6 and N7 of the guanine in RTGR. The chelated Fe2+ is ionically attracted to the phosphate that lies 5' to the dG, causing it to push on the thymidine methyl, thereby flipping the thymidine out of the helix. Major groove compression results and the two adjacent adenine N7 residues are attracted to the iron, inducing an octahedral chelation site in the DNA. The end result is that the iron is chelated by the three purine N7 residues as well as by the guanine O6 and the phosphate group between the deoxyguanosine and thymidine in the ATGA sequence. B, a schematic of the modeled iron chelation site compared with that expected for an octahedral chelation site. Since    and the sum of , , plus   360°, the lone electron pairs of the three purine N7 atoms are oriented such that they could chelate an iron among them. C, a stick model of the computer-generated distorted DNA structure. The iron and sodium are not shown for clarity of viewing, and the thymidine deoxyribose is highlighted in brown. The complementary strand is blue; the nucleotides flanking the ATGA are yellow. D, a space-filling model of the simulation result. The Fe2+ is shown in dark gray; it is chelated to the three purine N7 residues and the phosphate between the dT and the dG, and it is juxtaposed to the thymidine deoxyribose highlighted in brown. Non-consensus nucleotides are colored as in C.

While these are predictions of a modeling exercise, a hypothetical scenario is clear. Fe²⁺ is attracted to the polyanionic DNA molecule and moves along the phosphate backbone among individual phosphate residues. It is attracted weakly by one or two purine N⁷ residues, but if in a RTGR sequence the dT moves (or had moved) aside and then the Fe²⁺ coordinates with the third purine N⁷, trapping it in a free-energy sink, at least long enough to show preferential cleavage at the RTGR sequence.

Using the energy minimization model as a working hypothesis, we monitored the effects of Fe²⁺ on the proton NMR signals of an ATGA-containing 16'-mer duplex. Iron was titrated into the DNA solution, and one-dimensional scans were taken (Fig. 6A). We noted iron-dependent changes in the chemical shifts of several imino region (12-14 ppm) and thymidine methyl region (0.5-2 ppm) peaks (Fig. 6, A and B). (Iron-dependent changes also were noted in the aromatic region notably in the T8 aromatic peak; data not shown.) Changes in the observed chemical shifts are centered around the ATGA sequence, and taper off toward the ends of the oligomer (Fig. 6B). To see if these changes were due to Fe²⁺ as opposed to Fe³⁺, we allowed the Fe²⁺ to oxidize. The observed changes in chemical shift reversed during a 16-h incubation under an aerobic atmosphere. Under nitrogen, however, the changes in chemical shift were stable (data not shown). Since the energy minimization model predicted that the thymine methyl in RTGR is a necessary trigger for this interaction, we tested this interaction with an oligonucleotide that contained an identical DNA sequence, except that the thymine in RTGR was replaced by a uracil (Fig. 6C). Clearly the broadening and changes in chemical shift seen for RTGR in Fig. 6 (A and B) did not occur in the presence of Fe²⁺. Hence, the thymine methyl plays a necessary role in the iron binding as predicted by the model.

View larger version (15K): [in this window] [in a new window]   Fig. 6.   Effect of Fe2+ on the 1H NMR methyl peaks for 16'-mer duplexes containing ATGA or related sequences. A-D, the NaCl concentration was 130 mM and the solvent was water with 10% D2O. FeSO4 was mixed with the DNA at a ratio of Fe2+ to duplex oligonucleotide of 0.33 or 0.36. The one-dimensional 1H NMR scans were taken at 22 °C on a Bruker DRX 500 MHz spectrometer at a proton frequency of 500 MHz. Water suppression was achieved with a jump-and-return pulse or using a WATERGATE sequence with a gradient probe. Data were processed on SGI computers using Bruker software. A, the duplex consisted of CGCGATATGACACTAG hybridized to CTAGTGTCATATCGCG and was present at 1.1 mM. 65,536 complex points were taken in the time domain using a spectral width of 15 kHz, and 128 scans were taken for each one-dimensional spectrum. To make assignments, the chemical shifts of all non-exchangeable protons, except for the 5' protons, had been determined by two-dimensional NOESY in D2O in the absence of Fe2+. The chemical shifts of all the imino and cytosine amino protons, except for those at the terminal base pairs, had been determined by two-dimensional NOESY in H2O. The chemical shifts of the methyl protons due to the presence of Fe2+ were as follows: 1.60 (T18), 1.50 (T14), 1.37 (T26), 1.27 (T8), 1.27 (T6), 1.14 (T28), 1.13 (T21), and 0.94 (T23). The nucleotide numbering scheme is shown in B. B, the iron-dependent changes in chemical shift of the thymidine methyl protons in the RTGR-containing oligomer are calculated from A. From the same two spectra, the changes in chemical shift were calculated for the imino protons. The cross-hatched bars represent the absolute change in shift of the methyl protons (0.5-2 ppm), and the solid bars represent the change in shift of the imino protons (12-14 ppm). C, CGCGATAUGACACTAG was hybridized to CTAGTGTCATATCGCG to form an oligomer that is the same as that in A except that uracil was substituted for the thymine in the ATGA portion. 32,768 complex points were taken in the time domain using a spectral width of 10 kHz, and 256 scans were taken for each one-dimensional spectrum. The assignments of the methyl protons in the RUGR-containing sequence were deduced by comparing peaks with those of the RTGR-containing sequence. The duplex oligomer concentration was 0.26 mM. The absence of iron-dependent peak shifting was also observed using 1.5 mM duplex. D, CGCGATA5mCGACACACTAG was hybridized to CTAGTGTCGTATCGCG to form an oligomer that is the same as that in A, except that 5-methyl cytosine was substituted for thymine in the ATGA portion and, on the opposing strand, a corresponding guanine was introduced in lieu of the adenine. 32,768 complex points were taken in the time domain using a spectral width of 10 kHz, and 128 scans were taken for each one-dimensional spectrum. To make assignments, the chemical shifts of all non-exchangeable protons, except for the 5' protons were determined by two-dimensional NOESY in D2O in the absence and the presence of Fe2+. The chemical shifts of the methyl protons due to the presence of Fe2+ were as follows: 1.63 (T18), 1.52 (T14), 1.44 (5mC), 1.36 (T26), 1.28 (T6), 1.20 (T28), 1.18 (T21), and 0.92 (T23).

To further test which aspects of the thymine base might be necessary for interaction with Fe²⁺, we substituted the thymine in RTGR with a 5-methylcytosine (^5mC) and on the opposing strand we introduced a corresponding guanine in lieu of the adenine. In this case a pattern of chemical shifts similar to that seen with the substrate containing RTGR was again evident (Fig. 6D). Notably, the ^5mC methyl resonance was shifted downfield by 0.1 ppm. We can also compare the Fe²⁺-dependent change in chemical shift of the T8 methyl group to published differences in chemical shifts between stacked and extrahelical thymidine residues. Morden et al. (35) showed that the thymine methyl proton peak of a bulged extrahelical thymidine is 0.05-0.2 ppm downfield to the methyl proton peaks of all the paired and stacked thymidines. Moreover, Kalnik et al. (36) showed that when a bulged and unpaired thymine goes from a predominantly stacked to a predominantly extrahelical conformation, the methyl resonance of this thymine shifts downfield by up to 0.2 ppm. This latter unpaired thymine was part of an RTGR sequence (GTGA), so it is in an environment similar to the thymine of the ATGA sequence of our duplex oligonucleotide. In our case, the T8 (or the ^5mC8) resonance was uniquely downfield shifted by ~0.1 ppm upon addition of Fe²⁺. Hence, the observed shifts in the presence of Fe²⁺ are consistent with the target pyrimidines becoming extrahelical as predicted by the simulation.

It must be noted that in Fig. 6A there was no superimposition of spectra due to the presence of two species of DNA, even though substoichiometric amounts of Fe²⁺ were used. Moreover, we have noted that there is continuous shifting and broadening of resonances upon increasing additions of iron (data not shown). Therefore, it seems that compared with the chemical shift time scale for NMR, the iron is in fast exchange among these small duplexes. We had predicted that type I oxidants arise on Fe²⁺ ions, which are loosely bound to DNA (26), and loose binding would give rise to fast exchange. Alternately, the rapidity of exchange that we observed might reflect the short length of the oligonucleotide, and other physical means are called for to measure the stability of the complex within longer, possibly circular DNAs.

	DISCUSSION

Top Abstract Introduction Procedures Results Discussion References

A definitive conclusion of this study is that different sequences in duplex DNA undergo preferential nicking by Fe²⁺ in 0.5 mM versus 50 mM H₂O₂. This result is consistent with the proposal that the different kinetics of DNA damage by Fe²⁺ and H₂O₂ at these two H₂O₂ concentrations is attributable to radicals formed on iron ions that interact differently with DNA and hence have different properties. What was not anticipated, however, was the adherence of preferential nicking to a small number of consensus sequences which differed for the two H₂O₂ concentrations. It is unlikely that ^·OH (or a similar oxidant) could distinguish between different sequences of 4 bp. Thus we presume that the preferential cleavages observed are at least partially due to preferential iron binding at consensus sequences, although they might also be due to an enhanced ability of Fe²⁺ to react with H₂O₂ when so bound. Indeed, electronegativity is not equally distributed along a DNA helix but is dependent upon the local DNA sequence, and the binding of transition metals to DNA is thought to involve association with phosphate residues followed by transfer to preferred sequence regions to form a complex whose stability depends upon the particular metal ion, the DNA sequence, and the secondary structure of the DNA in the region (37-41). It should be emphasized, however, that the location of DNA nicks tells us where oxidants attacked the deoxyribose residues in DNA. The next phase of these studies will be to look at which sequences might give rise to particular base damages, e.g. 8-oxoguanine, thymine glycol, etc. Only after such studies will the full spectrum of Fe²⁺ binding sites on DNA begin to emerge.

Many authors have pointed out that the strongest base component for chelation of transition metals on DNA is the purine N⁷ and that regions around a G-rich site are especially electronegative around the N⁷ atoms because of strong stacking interactions (39, 40, 42-44). The binding of iron to adjacent G residues would not be hindered by the exocyclic amino hydrogens of adenine and would possibly be supported by guanine C⁶ oxygens. The complex might also be stabilized by phosphate residues on the same strand which are 5' to the respective dG residues (42). Such an association with a guanine N⁷ and the 5'-proximal phosphate would place the Fe²⁺ in close proximity of the deoxyribose moiety 5' to the dG. This model correlates nicely with the observed cleavage pattern in GGG and with the decrease in ionization potential in the 3' to 5' direction (38, 44). Furthermore, GGG sequences might form sinks for oxidizing radicals (30, 44). The resistance of cleavages at RGGG sequences to ethanol and H₂O₂ might reflect the dispersion of a radical electron among the stacked guanines, thus shielding it from exogenous agents. It is noteworthy that copper/ascorbate/5 mM H₂O₂ (17) and Zn²⁺ (16) have recently been reported also to interact with DNA at runs of dGMP. In the future we plan to study the interaction of Fe²⁺ with the type II radical target, RGGG. These studies are especially interesting because of the presence of RGGG in a the large majority of telomere repeats. Should damage to DNA in vivo reflect what was observed with the plasmid studies, one might speculate that this consensus sequence in telomeres was selected to sequester transition metals which have been attracted to DNA in order to protect other sequences. However, this preferential binding could also contribute to H₂O₂-dependent telomere shortening and a reduction of the "Hayflick limit," which is observed in model systems of aging (45, 46).

In the RTGR consensus sequence, a Fe²⁺-DNA interaction in the minor groove would either place the iron too far from the guanine or too far from the preferentially cleaved sugar. On the other hand, if the iron were adjacent to the deoxyribose of thymidine in the major groove, the iron could also be within Van der Waals distance to the guanine N⁷. For sequences of more than two base pairs to interact with iron in the minor groove in the vicinity of the preferentially cleaved deoxyribose, the DNA must be distorted to the extent that it does not resemble B-DNA. On the other hand, with only small distortions of the B-structure of DNA, a Fe²⁺ ion in the major groove could access functional groups from 3 base pairs and still be in the vicinity of the preferentially cleaved deoxyribose moiety and chelating phosphates. Just as with protein-DNA interactions, it seems that metal-DNA sequence-specific interactions are more likely to occur in the major groove than in the minor groove (22).

As shown by NMR, the interaction of Fe²⁺ with the RTGR-containing 16'-mer duplex is readily reversible. This reversibility offers an explanation for the sensitivity of type I oxidations to high H₂O₂ concentrations. The Fe²⁺ might be oxidized by H₂O₂ while it is free, and the oxidant produced by this Fenton reaction quenched by H₂O₂ before damaging DNA. However, when formed within the RTGR sequence, the oxidant reacts immediately with DNA rather than with H₂O₂. One might then predict that Fe²⁺ interactions with RGGG would not be so readily reversible, as in this case damages are not quenched by H₂O₂ or ethanol.

The sequence RTGR occurs in several biologically interesting regions of DNA. ATGG is part of the human and Drosophila centromeric repeats, AATGG, and the yeast repeat is AAATGG (47). Moreover, the RTGR motif seems to occur as a required element in the regulatory regions of genes that respond to iron- or oxidative stress. The consensus sequence of the "iron box," a key promoter element involved in iron homeostasis of E. coli, has two RTGR sequences in dyad symmetry (inverted repeats) (48). Two RTGR sequences in dyad symmetry are also found in three consensus sequence elements in the promoter that regulates the major mammalian AP endonuclease activity which is necessary for the repair of the oxidatively damaged DNA (49). FRE1 is a component of the principal yeast iron uptake system. The 5' non-coding segment of FRE1 that is necessary for conferring iron-regulatory transcriptional activity contains RTGR in repeated GRTGAGCAAAAA sequences (50). It has been shown that the RTGR motif is directly involved in the regulation of certain chemoprotective responses mediated by the antioxidant-responsive element (ARE) (51), and it is interesting to note that H₂O₂ induces gene activity under the control of the ARE element. For example, the ARE that is found in the 5'-flanking regions of the rat and mouse glutathione S-transferase Ya subunit genes, the rat and human NAD(P)H:quinone reductase genes, and -glutamyl cysteine synthetase genes each contains three RTGR sequences. In deletion studies by Rushmore et al. (51), all six of the deletions that infringe upon the RTGR motifs cause significant reductions (>33%) in the basal and/or inducible promoter activity. Only once in 13 other deletions was this response affected to the same degree (51). Further studies of the ARE consensus sequence showed that substitution of the dT in the core RTGR sequence resulted in complete loss of inducibility by chemoprotective inducers (52, 53). The studies of Figs. 5 and 6 have implied a similar requirement for dT in the specific interaction of RTGR with Fe²⁺, and a similar interaction is not seen between RTGR and Fe³⁺. Perhaps these regulatory regions are able to sense the oxidative environment of the cell via interactions of their RTGR motifs with reduced iron.

We are currently determining the three-dimensional structure of the interaction of Fe²⁺ with the small oligonucleotide complex by two-dimensional NOESY. Once the structure is known, we plan to study several of the naturally occurring regulatory motifs both physically and biochemically in order to learn why this iron-binding sequence is required in these regulatory motifs.

	ACKNOWLEDGEMENTS

We thank Matthew S. Falk for helping devise the PM2 DNA restriction scheme, and Corey Liu, Mark Kubinec, David Wemmer, and the University of California Berkeley College of Chemistry NMR facility for the use of the NMR machines and their assistance. The NMR machines are maintained by Instrumentation Grants DE FG05-86ER75281 from the U.S. Department of Energy and DMB 86-09305 and BBS 87-20134 from the National Science Foundation.

	FOOTNOTES

^* This work was supported by National Institutes of Health Grants R37GM19020, T32ES070075, and P30ES01896.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Present address: Sepracor, Inc., Marlborough, MA 01752.

^§ Present address: Dept. of Pharmaceutical Chemistry, University of California, San Francisco, CA 94143.

^¶ Present address: Kosan Biosciences, Inc., Burlingame, CA 94010.

To whom correspondence should be addressed. Tel.: 510-642-7583; Fax: 510-643-5035; E-mail: slinn{at}socrates.berkeley.edu.

The abbreviations used are: HPLC, high performance liquid chromatography; Y, pyrimidine; R, purine; N, any of the four DNA bases; bp, base pair(s); ARE, antioxidant-responsive element; TBE, Tris borate-EDTA; NOESY, nuclear Overhauser effect spectroscopy.

	REFERENCES

Top Abstract Introduction Procedures Results Discussion References

1. Henle, E. S., and Linn, S. (1997) J. Biol. Chem. 272, 19095-19098[Free Full Text]
2. Imlay, J. A., Chin, S. M., and Linn, S. (1988) Science 240, 640-642[Abstract/Free Full Text]
3. Imlay, J. A., and Linn, S. (1988) Science 240, 1302-1309[Abstract/Free Full Text]
4. Walling, C. (1975) Accts. Chem. Res. 8, 125-131[CrossRef]
5. Yamazaki, I., and Piette, L. H. (1991) J. Am. Chem. Soc. 113, 7588-7593[CrossRef]
6. Goldstein, S., Meyerstein, D., and Czapski, G. (1993) Free Radical Biol. Med. 15, 435-445[CrossRef][Medline] [Order article via Infotrieve]
7. Koppenol, W. H. (1994) in Free Radical Damage and Its Control (Rice, C. A., and Burdon, R. H., eds), pp. 3-24, Elsevier Science, New York
8. Wink, D. A., Nims, R. W., Saavedra, J. E., Utermahlen, W. E., Jr., and Ford, P. C. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 6604-6608[Abstract/Free Full Text]
9. Luo, Y., Han, Z. X., Chin, M., and Linn, S. (1994) Proc. Natl. Acad. Sci., U. S. A. 91, 12438-12442[Abstract/Free Full Text]
10. Kaneko, M., Kodama, M., and Inoue, F. (1994) Free Radical Res. 20, 229-239[Medline] [Order article via Infotrieve]
11. Bertoncini, C. R. A., and Meneghini, R. (1995) Nucleic Acids Res. 23, 2995-3002[Abstract/Free Full Text]
12. Chevion, M. (1988) Free Radical Biol. Med. 5, 27-37[CrossRef][Medline] [Order article via Infotrieve]
13. Clayson, D. B., Mehta, R., and Iverson, F. (1994) Mutat. Res. 317, 25-42[Medline] [Order article via Infotrieve]
14. Meneghini, R. (1997) Free Radical Biol. Med. 23, 783-792[CrossRef][Medline] [Order article via Infotrieve]
15. Enright, H. U., Miller, W. J., and Hebbel, R. P. (1992) Nucleic Acids Res. 20, 3341-3346[Abstract/Free Full Text]
16. Martínez-Balbás, M. A., Jiménez-García, E., and Azorín, F. (1995) Nucleic Acids Res. 23, 2464-2471[Abstract/Free Full Text]
17. Rodriguez, H., Drouin, R., Holmquist, G. P., O'Connor, T. R., Boiteux, S., Laval, J., Doroshow, J. H., and Akman, S. A. (1995) J. Biol. Chem. 270, 17633-17640[Abstract/Free Full Text]
18. Samuni, A., Aronovitch, J., Godinger, D., Chevion, M., and Czapski, G. (1983) Eur. J. Biochem. 137, 119-124[Medline] [Order article via Infotrieve]
19. Mello-Filho, A. C., and Meneghini, R. (1991) Mutat. Res. 251, 109-113[Medline] [Order article via Infotrieve]
20. Keyer, K., and Imlay, J. A. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 13635-13640[Abstract/Free Full Text]
21. Oikawa, S., and Kawanishi, S. (1996) Biochemistry 35, 4584-4590[CrossRef][Medline] [Order article via Infotrieve]
22. Dervan, P. B. (1986) Science 232, 464-590[Abstract/Free Full Text]
23. Absalon, M. J., Kozarich, J. W., and Stubbe, J. (1995) Biochemistry 34, 2065-2075[CrossRef][Medline] [Order article via Infotrieve]
24. Henner, W. D., Granburg, S. M., and Haseltine, W. A. (1982) J. Biol. Chem. 257, 11750-11754[Abstract/Free Full Text]
25. Luo, Y., Henle, E. S., and Linn, S. (1996) J. Biol. Chem. 271, 21167-21176[Abstract/Free Full Text]
26. Luo, Y., Henle, E. S., Chattopadhyaya, R., Jin, R., and Linn, S. (1994) Methods Enzymol. 234, 51-59[CrossRef][Medline] [Order article via Infotrieve]
27. Maxam, A. M., and Gilbert, W. (1977) Proc. Natl. Acad. Sci, U. S. A. 74, 560-564[Abstract/Free Full Text]
28. Plum, G. E., Grollman, A. P., Johnson, F., and Breslauer, K. J. (1995) Biochemistry 34, 16148-16160[CrossRef][Medline] [Order article via Infotrieve]
29. Kuhnlein, U., Penhoet, E. E., and Linn, S. (1976) Proc. Natl. Acad. Sci. U. S. A. 73, 1169-1173[Abstract/Free Full Text]
30. Hall, D. B., Holmlin, R. E., and Barton, J. K. (1996) Nature 382, 731-735[CrossRef][Medline] [Order article via Infotrieve]
31. Henderson, E. (1995) in Telomeres (Blackburn, E., and Greider, C. W., eds), pp. 11-34, Cold Spring Harbor Press, Cold Spring Harbor, NY
32. Roberts, R. J. (1995) Cell 82, 9-12[CrossRef][Medline] [Order article via Infotrieve]
33. Slupphaug, G., Mol, C. D., Kavali, B., Arvai, A. S., Krokan, H. E., and Tainer, J. A. (1996) Nature 384, 87-92[CrossRef][Medline] [Order article via Infotrieve]
34. Terrón, A. (1993) Comments Inorg. Chem. 14, 63-88
35. Morden, K. M., Gunn, B. M., and Maskos, K. (1990) Biochemistry 29, 8835-8845[CrossRef][Medline] [Order article via Infotrieve]
36. Kalnik, M. W., Norman, D. G., Li, B. F., Swann, P. F., and Patel, D. J. (1990) J. Biol. Chem. 265, 636-647[Abstract/Free Full Text]
37. Pezzano, H., and Podo, F. (1980) Chem. Rev. 5, 366-401
38. Pullman, A., and Pullman, B. (1981) Q. Rev. Biophys. 14, 289-380[Medline] [Order article via Infotrieve]
39. Mattes, W. B., Hartley, J. A., and Kohn, K. W. (1986) Nucleic Acids Res. 14, 2971-2987[Abstract/Free Full Text]
40. Sagripanti, J. L., and Kraemer, K. H. (1989) J. Biol. Chem. 264, 1729-1739[Abstract/Free Full Text]
41. Hartley, J. A., Forrow, S. M., and Souhami, R. L. (1990) Biochemistry 29, 2985-2991[CrossRef][Medline] [Order article via Infotrieve]
42. Sissoeff, I., Grisvard, J., and Guille, E. (1976) Prog. Biophys. Mol. Biol. 31, 165-199[Medline] [Order article via Infotrieve]
43. Saenger, W. (1984) Principles of Nucleic Acid Structure, pp. 201-219, Springer Verlag, Inc., New York
44. Saito, I., Takayama, M., Sugiyama, H., and Nakatani, K. (1995) J. Am. Chem. Soc. 117, 6406-6407[CrossRef]
45. Chen, Q., and Ames, B. N. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 4130-4134[Abstract/Free Full Text]
46. Petersen, S., Saretzki, G., and von Zglinicki, T. (1998) Exp. Cell Res. 239, 152-160[CrossRef][Medline] [Order article via Infotrieve]
47. Grady, D. L., Ratliff, R. L., Robinson, D. L., McCanlies, E. C., Meyne, J., and Moyzis, R. K. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 1695-1699[Abstract/Free Full Text]
48. Neilands, J. B. (1995) J. Biol. Chem. 270, 26723-26726[Free Full Text]
49. Harrison, L., Ascione, A. G., Takiguchi, Y., Wilson, D. M., III, Chen, D. J., and Demple, B. (1997) Mutat. Res. 385, 159-172[Medline] [Order article via Infotrieve]
50. Dancis, A., Roman, D. G., Anderson, G. J., Hinnebusch, A. G., and Klausner, R. D. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 3869-3873