Oxygen-dependent DNA damage amplification involving 5,6-dihydrothymidin-5-yl in a structurally minimal system

Article abstract of DOI:10.1021/ja010180s

5,6-Dihydrothymidin-5-yl (1) was independently generated in a dinucleotide from a phenyl selenide precursor (4). Under free radical chain propagation conditions, the products resulting from hydrogen atom donation and radical-pair reaction are the major ob

Full text of DOI:10.1021/ja010180s

J. Am. Chem. Soc. 2001, 123, 5181-5187
5181
Oxygen-Dependent DNA Damage Amplification Involving
5,6-Dihydrothymidin-5-yl in a Structurally Minimal System
Keri A. Tallman and Marc M. Greenberg*
Contribution from the Department of Chemistry, Colorado State UniVersity, Fort Collins, Colorado 80523
ReceiVed January 22, 2001
Abstract: 5,6-Dihydrothymidin-5-yl (1) was independently generated in a dinucleotide from a phenyl selenide
precursor (4). Under free radical chain propagation conditions, the products resulting from hydrogen atom
donation and radical-pair reaction are the major observed products in the absence of O2. The stereoselectivity
of the trapping process is dependent on the structure of the hydrogen atom donor. No evidence for
internucleotidyl hydrogen atom abstraction by 1 was detected. The tandem lesion (17) resulting from hydrogen
atom abstraction from the C1′ position of the adjacent 2′-deoxyuridine by the peroxyl radical derived from 1
(3) is observed under aerobic conditions. The structure of this product is confirmed by independent synthesis
and its transformation into a second independently synthesized product (24). Internucleotidyl hydrogen atom
abstraction is effected selectively by the 5S-diastereomer of the peroxyl radical. The formation of dinucleotide
17 provides further support for the novel O2-dependent DNA damage amplification mechanism involving 1
reported previously (Greenberg, M. M.; et al. J. Am. Chem. Soc. 1997, 119, 1828).
Nucleic acids are oxidatively damaged by a variety of natural
and unnatural products. Some of these molecules, such as the
neocarzinostatin chromophore and members of the enediyne
family of antitumor antibiotics, are capable of producing
bistranded lesions due to the biradical nature of their reactive
form.1 Additionally, a single molecule of bleomycin is capable
of forming multiple lesions via recycling of its redox-active
metallocenter.2 Investigations of DNA damage mediated by
γ-radiolysis brought to light the possibility that a single initial
nucleic acid damaging event could yield bistranded and tandem
lesions.3,4 Currently, there is significant interest in the possible
role of these types of lesions in DNA repair and mutagenesis.5
Using γ-radiolysis studies as a guide, we carried out chemical
studies, from which it was concluded that the nucleobase radical
5,6-dihydrothymidin-5-yl (1)6 produces tandem lesions in oli-
godeoxynucleotides via an O2-dependent DNA damage ampli-
fication mechanism (Scheme 1).7,8 We now wish to report
product studies employing the minimal chemical unit capable
of sustaining such a process, a dinucleotide (4), which support
this nucleic acid strand damage mechanism. Analysis of pho-
tolysis experiments on 4 by HPLC using independently syn-
thesized potential products as standards corroborates the conclu-
sions reached previously and provides additional details about
this chemical process (eq 1).
(1) Xi, Z.; Goldberg, I. H. In ComprehensiVe Natural Products Chemistry,
Volume 7, DNA and Aspects of Molecular Biology; Kool, E. T., Ed.;
Pergamon: Oxford, 1999; pp 553-592.
(2) (a) Steighner, R. J.; Povirk, L. F. Proc. Natl. Acad. Sci. U.S.A. 1990,
87, 8350. (b) Absalon, M. J.; Kozarich, J. W.; Stubbe, J. Biochemistry 1995,
34, 2065. (c) Absalon, M. J.; Wu, W.; Kozarich, J. W.; Stubbe, J.
Biochemistry 1995, 34, 2076.
(3) (a) von Sonntag, C. The Chemical Basis of Radiation Biology; Taylor
and Francis, London, 1987. (b) Greenberg, M. M. In ComprehensiVe Natural
Products Chemistry, Volume 7, DNA and Aspects of Molecular Biology,
Kool, E. T., Ed.; Pergamon: Oxford, 1999; pp 371-426. (c) Greenberg,
M. M. Chem. Res. Toxicol. 1998, 11, 1235.
(4) Aslam Siddiqi, M.; Bothe, E. Radiat. Res. 1987, 112, 449.
(5) (a) Sutherland, B. M.; Bennett, P. V.; Sidorkina, O.; Laval, J. Proc.
Natl. Acad. Sci. U.S.A. 2000, 97, 103. (b) Satoh, M. S.; Jones, C. J.; Wood,
R. D.; Lindahl, T. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 6335.
(6) For simplicity, molecules are referred to by the same number whether
they are present as a monomeric nucleoside or as part of a biopolymer.
(7) (a) Barvian, M. R.; Greenberg, M. M. J. Org. Chem. 1995, 60, 1916.
(b) Barvian, M. R.; Greenberg, M. M. J. Am. Chem. Soc. 1995, 117, 8291.
(c) Greenberg, M. M.; Barvian, M. R.; Cook, G. P.; Goodman, B. K.;
Matray, T. J.; Tronche, C.; Venkatesan, H. J. Am. Chem. Soc. 1997, 119,
1828.
5,6-Dihydrothymidin-5-yl (1) is a reactive intermediate that
is formed in significant amounts upon exposure of DNA to
γ-radiolysis.3,9,10 This nucleobase radical is produced as a
consequence of the indirect effect of ionizing radiation via
hydrogen atom addition to the pyrimidine double bond of
thymidine, and represents the major reactive intermediate formed
as a result of the reaction between these two entities. The
analogous hydroxyl radical adduct (5) is also produced in
significant amounts. In addition, the favorable reduction potential
of thymidine, compared to those of the other three native
nucleotides in DNA, results in the formation of 1 via sequential
electron addition and protonation as perhaps the most abundant
(9) (a) Symons, M. C. R. J. Chem. Soc., Faraday Trans. 1 1987, 83, 1.
(b) Boon, P. J.; O’Connell, A.; Podmore, I. D.; Symons, M. C. R. J. Chem.
Soc., Perkin Trans. 2 1993, 2061. (c) Cullis, P. M.; Langman, S.; Podmore,
I. D.; Symons, M. C. R. J. Chem. Soc., Faraday Trans. 1990, 86, 3267.
(10) Cullis, P. M.; McClymont, J. D.; Malone, M. E.; Mather, A. N.;
Podmore, I. D.; Sweeney, M. C.; Symons, M. C. R. J. Chem. Soc., Perkin
Trans 2 1992, 1695.
(8) (a) Deeble, D. J.; von Sonntag, C. Int. J. Radiat. Biol. 1986, 49,
927. (b) Schulte-Frohlinde, D.; Bothe, E. Z. Naturforsch. 1984, 39c, 315.
10.1021/ja010180s CCC: $20.00 © 2001 American Chemical Society
Published on Web 05/11/2001

5182 J. Am. Chem. Soc., Vol. 123, No. 22, 2001
Tallman and Greenberg
Scheme 1
Scheme 2
reactive intermediate produced from thymidine as a consequence
of this two-step process. 5,6-Dihydrothymidin-5-yl (1) has been
suggested to play a significant role in nucleic acid strand
damage. It was proposed that 1 amplifies the initial strand
damage event (its formation) by directly abstracting a hydrogen
atom from the nucleotide bonded to its 5′-phosphate, resulting
in the formation of tandem lesions.9
opposed to an alkoxyl radical, which may result from reduction
of 3. Alkoxyl radicals are expected to be far more promiscuous
in their hydrogen atom abstraction reactions from carbon-
hydrogen bonds.14,15 Later studies on monomers and biopoly-
mers, in which the C1′ radical of 2′-deoxyuridine (7) was
independently generated, corroborated the O2-dependent DNA
damage amplification mechanism (Scheme 2).16-18 In addition,
further investigation of the reactivity of 1 at the monomeric
level under aerobic conditions revealed the existence of a
competing pathway for the peroxyl radical 3 that reconstitutes
the native nucleoside with concomitant release of superoxide
(eq2).17 Although these reports provide significant information,
Chemical studies in which 1 was generated from a ketone
precursor (2) via Norrish type I photocleavage were inconsistent
with such a proposal. At the monomeric level, 1 failed to abstract
hydrogen atoms from molecules chosen to mimic the carbohy-
drate components of nucleosides.6 Furthermore, direct strand
breaks or alkali-labile lesions were not observed when 1 was
produced from 2 under anaerobic conditions in single-stranded
oligodeoxynucleotides.7b,c The consequences of generating 1
from 2 under aerobic conditions were considerably different.
At the monomeric level, a diastereomeric mixture of the unstable
tertiary hydroperoxide (6) was isolated.7a When produced in
single-stranded oligodeoxynucleotides under aerobic conditions,
1 gave rise to alkali-labile lesions, and to a lesser extent direct
strand breaks, predominantly at the nucleotide bonded to the
5′-phosphate of 1. Furthermore, the initial site of 1 was
transformed into an alkali-labile lesion, indicating that two
lesions are formed from one initial interaction between DNA
and ionizing radiation. This mechanism for O2-dependent DNA
damage amplification represented a pathway that was distinct
from any other DNA-damaging pathway documented in the
chemical literature.
A variety of experiments, including studies of kinetic
(product) isotope effects and enzymatic end group analysis, led
to a proposed mechanism involving selective C1′ hydrogen atom
abstraction by the nucleobase peroxyl radical derived from 1
(Scheme 1).7 No evidence (as indicated by the lack of an isotope
effect upon deuteration of the C4′ position or the formation of
3′-phosphoglycolate termini) for competing hydrogen atom
abstraction from the slightly stronger carbon-hydrogen bond
at the C4′ position was obtained.3b,11,12 This selectivity is
consistent with that expected for a peroxyl radical (e.g., 3), as
a number of issues regarding this DNA damage amplification
(11) Kozarich, J. W.; Worth, L., Jr.; Frank, B. L.; Christner, D. F.;
Vanderwall, D. E.; Stubbe, J. Science 1989, 245, 1396.
(12) (a) Stubbe, J.; Kozarich, J. W. Chem. ReV. 1987, 87, 1107. (b) Dussy,
A.; Meggers, E.; Giese, B. J. Am. Chem. Soc. 1998, 120, 7399.
(13) (a) Colson, A.-O.; Sevilla, M. D. J. Phys. Chem. 1995, 99, 3867.
(b) Miaskiewicz, K.; Osman, R. J. Am. Chem. Soc. 1994, 116, 232.
(14) Ho, W. F.; Gilbert, B. C.; Davies, M. J. J. Chem. Soc., Perkin Trans.
2 1997, 2525.
(15) (a) Avilla, D. V.; Brown, C. E.; Ingold, K. U.; Lusztyk, J. J. Am.
Chem. Soc. 1993, 115, 466. (b) Weber, M.; Fischer, H. J. Am. Chem. Soc.
1999, 121, 7381.
(16) (a) Tronche, C.; Goodman, B. K.; Greenberg, M. M. Chem. Biol.
1998, 5, 263. (b) Hwang, J.-T.; Greenberg, M. M. J. Am. Chem. Soc. 1999,
121, 4311.
(17) Tallman, K. A.; Tronche, C.; Yoo, D. J.; Greenberg, M. M. J. Am.
Chem. Soc. 1998, 120, 4903.

Oxygen-Dependent DNA Damage Amplification
J. Am. Chem. Soc., Vol. 123, No. 22, 2001 5183
Scheme 3a
Scheme 4a
a Reagents: (a) 10% TFA/H2O, THF, 0 °C; (b) Et3N‚3HF, THF.
the ultimate radical precursor was facilitated by using single
diastereomers of 9. We also wanted to independently determine
that the stereochemistry of the phenyl selenide precursor 4 did
not affect the observed products. The diastereomers of 9 were
separated as their respective 3′-O-tert-butyldimethylsilyloxy
compounds (12), which were prepared from the previously
reported bis-silyl protected substrate (11) using aqueous TFA
in THF (Scheme 4).20 Following removal of the remaining silyl
group, the fully deprotected phenyl selenides 9 were coupled
to the â-cyanoethyl phosphoramidite of 2′-deoxyuridine. Prod-
ucts corresponding to coupling at the 3′ hydroxyl group were
not observed. Following cleavage of the â-cyanoethyl group
from a diastereomeric mixture of phosphate triesters, the desired
products were obtained as their sodium salts (Dowex Na+).
a Reagents: (a) 2′-deoxyuridine â-cyanoethyl phosphoramidite,
tetrazole/CH3CN (0.5 M), THF, 0 °C; 1.0 M I2, THF/2,6-lutidine/H2O
(2:2:1 by volume); (b) 50% Et3N/THF, reflux.
process remain unresolved. This collection of independent
experiments did not provide any information on the competition
(and hence efficiency of the internucleotidyl hydrogen atom
transfer) between superoxide elimination and hydrogen atom
abstraction by 3. Studies using oligodeoxynucleotides did not
provide detailed product analysis in support of the proposed
mechanism (Scheme 1). Finally, experiments involving biopoly-
mers also did not rule out the possibility that 1 abstracts a
hydrogen atom from an adjacent nucleotide, but did show that
this event does not result in the formation of an alkali-labile
lesion or direct strand break in the absence of O2. These issues
are addressed in the studies described below.
Results and Discussion
Design, Synthesis, and Analysis of the Dinucleotide and
Possible Products Derived from 5,6-Dihydrothymidin-5-yl
(1). To more fully characterize the reactivity of 1 and its
respective peroxyl radical (3) with an adjacent nucleotide in an
oligonucleotide, we reduced the system to the minimal structural
entity that could sustain such a reaction process, a dinucleotide
(4). The dinucleotide 4 was designed with a dimethoxytrityl
group at its 5′-terminus in order to facilitate synthesis and
purification, as well as to enhance detection of the starting
material and products by UV absorption. 2′-Deoxyuridine was
chosen as the 5′-nucleotide relative to the radical center (instead
of thymidine) in order to distinguish the possible formation of
thymidine via 1 in experiments where the photolysates are
subjected to enzymatic digestion. The phenyl selenide 4 was
chosen as a precursor for 1 due to its greater efficiency for
photochemical generation of the radical than the isopropyl
ketone 2.7b,17
Initially, the diastereomeric dinucleotides ((5S)-,(5R)-4) were
synthesized individually via standard phosphoramidite coupling
methods (Scheme 3), although our previous studies had shown
that a diastereomeric mixture of the ketone radical precursor 2
did not adversely affect mechanistic studies of 1 and 3.7b,19
However, characterization of the synthetic intermediates and
Solubility limitations necessitated that the anticipated products
(13-17) be synthesized by coupling the respective 3′-silyloxy
monomers (19-21) to either 18 or the â-cyanoethyl phosphora-
midite of 2′-deoxyuridine. The requisite 3′-O-tert-butyldimeth-
(19) (a) Beaucage, S. L.; Caruthers, M. H. Tetrahedron Lett. 1981, 22,
(18) (a) Chatgilialoglu, C.; Ferreri, C.; Bazzanini, R.; Guerra, M.; Choi,
S.-Y.; Emanuel, C. J.; Horner, J. H.; Newcomb, M. J. Am. Chem. Soc.
2000, 122, 9525. (b) Emanuel, C. J.; Newcomb, M.; Ferreri, C.; Chatgil-
ialoglu, C. J. Am. Chem. Soc. 1999, 121, 2927.
1859. (b) Matteucci, M. D.; Caruthers, M. H. Tetrahedron Lett. 1980, 21,
719.
(20) Robins, M. J.; Sarker, S.; Samano, V.; Wnuk, S. F. Tetrahedron
1997, 53, 447.

5184 J. Am. Chem. Soc., Vol. 123, No. 22, 2001
Tallman and Greenberg
Table 1. Anaerobic Photolysis of Phenyl Selenide Dinucleotide
(4)a
ylsilyloxy nucleosides 19-21 were obtained via selective
desilylation of the bis-silyl-protected compounds as described
above for 12. 5,6-Dihydrothymidine (20) was coupled to the
2′-deoxyuridine phosphoramidite as a mixture of diastereomers
(5S:5R, 3:2), which were obtained via epimerization of the
dihydrothymidine mixture initially isolated upon hydrogenation
(5S:5R, 9:1). Similarly, the analogous dinucleotide containing
5,6-dihydro-5-hydroxythymidine (15) was prepared using a
diastereomeric mixture (5R:5S, 4:1) of the 3′-silyloxy derivative
of thymidine C5 hydrate (21). The desirability of using a mixture
of diastereomers enabled us to prepare the protected thymidine
C5 hydrate 21 by a more expedient route than that previously
reported.21 The C5 hydroxyl group was incorporated by quench-
ing the dianion used to prepare 11 with O2. The coupling of
the 3′-silyloxy-protected nucleosides 19-21 was carried out as
described above (Scheme 3). Following deprotection of the
mixture of phosphate triesters, the 3′-silyl groups were cleaved
from the dinucleotides using Et3N‚3HF. The workup of this
reaction was crucial in order to prevent detritylation of the dinu-
cleotides. Passing the crude reaction mixture through Amberlite
anion-exchange resin was determined to be the most suitable
method for neutralizing the reaction mixtures. Subsequent
chromatography and passage through a cation-exchange column
(Dowex Na+) yielded the desired products as their sodium salts.
yield (%)b
(5S)-13: mass balance
(5R)-13
trap
t-BuSH
â-mercaptoethanol 28 ( 3
13
38 ( 12 44 ( 9.0 1.5 ( 0.4
42 ( 2 0.6 ( 0.1
14
(%)b
82 ( 13
70 ( 1
a [4] ) 0.1 mM; [trap] ) 1.1 mM; solvent, CH3CN/H2O, 1:1 by
volume. b Reported yields and mass balances represent average yields
from individual experiments ( the standard deviation measured from
these averages.
solvent gradient was suitable for separating every product.
Consequently, samples were analyzed by reverse-phase HPLC
using two different gradients which had complementary advan-
tages. The samples were first analyzed using CH3CN as the
organic solvent. These conditions allowed separation of (5R)-
15, (5R)-13, and (5S)-13, as well as the diastereomers of 16
and 17. However, (5S)-15 and 14 coeluted under these condi-
tions. To determine the yields of (5S)-15 and 14, the samples
were analyzed with a second gradient program using MeOH as
the organic cosolvent. These conditions resulted in coelution
of (5R)-15 and (5S)-15, which were resolved from 14. Although
(5S)-15 was now separable from 14, both diastereomers of 16
and 17 coeluted with 14 under these conditions. Since the total
yield of 15 was determined using the MeOH gradient, and the
yield of (5R)-15 was determined from analysis under the
CH3CN conditions, the amount of (5S)-15 formed was derived
from the difference between these two measurements. Similarly,
once the yield of (5S)-15 was calculated, it could be subtracted
from the total yield of (5S)-15 and 14 that was obtained during
the analysis of the samples under the CH3CN conditions, pro-
viding the yield of the latter dinucleotide. In addition, conversion
of starting material (4) and the yield of 24, which is formed
upon elimination from 16 and 17 in the presence of N,N′-
dimethylethylenediamine, were determined using the CH3CN
gradient.
The lability of the 2-deoxyribonolactone (8) to alkaline
conditions required that the independent synthesis of dinucle-
otide products containing this lesion (16, 17) be carried out using
a latent form of the lactone.22 We and others have shown that
2-deoxyribonolactone can be produced photochemically from
nucleoside analogues that are stable to the conditions of
oligonucleotide synthesis/deprotection (Scheme 2).16-18,23,24 The
dinucleotides 22 and 23 containing the photochemical precursor
to the 2-deoxyribonolactone and the diastereomeric mixtures
of 5,6-dihydrothymidine (20) or thymidine C5 hydrate (21) were
prepared as described above for 13-15. The lactone-containing
dinucleotides 16 and 17 were generated by aerobic photolysis
(350 nm) of 22 and 23, respectively, and purified by flash
chromatography.
Photolysis of 4 under Anaerobic Conditions. Product yields
were found to depend strongly on the nature of the trap em-
ployed but, as expected, not on the stereochemistry at the C5
position of the phenyl selenide 4. Consequently, all subsequent
photolysis experiments employed a diastereomeric mixture of
4 which was enriched in the 5R-diastereomer (typically between
2:1 and 3:1, (5R)-:(5S)-4). When the dinucleotide 4 was
photolyzed in the presence of t-BuSH or â-mercaptoethanol
under degassed conditions, a significant amount of 14 was
observed in addition to the product resulting from hydrogen
atom donation to 1 (13, Table 1). The formation of 14 in the
presence of thiol is consistent with relatively inefficient radical
chain propagation (via reaction between thiyl radicals and 4)
when this trap is used, compared to Bu3SnH.17 Consequently,
a larger fraction of 4 is decomposed by direct photolysis in the
presence of thiol. This results in higher yields of thymidine-
containing product 14, which can be formed by reaction between
Although it was unnecessary to synthesize the diastereomers
of 13 and 15-17 separately, quantitation of each isomer
produced from the generation of 1 was desirable. No single
•
1 and the benzene selenyl radical (k2R , Scheme 5).
(21) Barvian, M. R.; Greenberg, M. M. J. Org. Chem. 1993, 58, 6151.
(22) (a) Kappen, L. S.; Goldberg, I. H. Biochemistry 1989, 28, 1027.
(b) Urata, H.; Yamamoto, K.; Akagi, M.; Hiroaki, H.; Uesugi, S.
Biochemistry 1989, 28, 9566.
(23) Kotera, M.; Bourdat, A. G.; Defrancq, E.; Lhomme, J. J. Am. Chem.
Soc. 1998, 120, 11810.
(24) Hwang, J.-T.; Tallman, K. A.; Greenberg, M. M. Nucleic Acids Res.
1999, 19, 3805.

Oxygen-Dependent DNA Damage Amplification
J. Am. Chem. Soc., Vol. 123, No. 22, 2001 5185
Scheme 5
kR
kinter
[R-dU]
[â-dU]
)
(3)
(
)
kR + kR
kH[RSH]
Further attempts were made to detect hydrogen atom abstrac-
tion from the C1′ position of 2′-deoxyuridine by 1 using
tetranitromethane as a mechanistic probe. Tetranitromethane is
known to react with alkyl radicals, and in particular with
reducing radicals such as 2′-deoxyuridin-1′-yl (7).18a,28 We
anticipated that if 7 was formed via hydrogen atom abstraction
by 1, reaction of tetranitromethane with the C1′ radical would
yield 2-deoxyribonolactone (8). Hence, internucleotidyl hydro-
gen atom abstraction by 1 might be expected to yield 16 in the
presence of tetranitromethane. However, 14 and 15 were the
only products observed when 4 was irradiated under anaerobic
conditions in the presence of this trapping agent. As discussed
above, the former product may have arisen from reaction
between the radical pair formed upon photolysis of 4. The
thymidine C5 hydrate-containing product 15 is believed to be
formed via trapping of 1 by tetranitromethane, which is
presumably much faster than the internucleotidyl process that
was being probed for, and is consistent with the reactivity of
other nucleobase radicals with nitroaromatic compounds.29
O2-Dependent DNA Damage Amplification by 1. Previous
The stereoselectivity of hydrogen atom donation was also
dependent upon the nature of the hydrogen atom donor. Reaction
of monomeric 1 with hydrogen atom donors produces 5,6-
dihydrothymidine as a 1:1 mixture of diastereomers.7a In
contrast, when 1 is generated in the dinucleotide, formation of
(5S)-13 is favored when t-BuSH is employed as a trap,
presumably because attack from the si face of 1 is hindered by
the 5′-O-dimethoxytrityl-2′-deoxyuridine nucleotide (Table 1).
The opposite stereoselectivity is observed when â-mercapto-
ethanol reacts with 1. Preferential attack from the more hindered
face to form (5R)-13 may be due to delivery of the hydrogen
atom donor to the si face of 1 via hydrogen bonding of the
thiol to the uracil moiety, and is discussed further below.
Probing Internucleotidyl Hydrogen Atom Abstraction by
1. Thermodynamic considerations led us to predict that if 1
abstracts hydrogen atom(s) from an adjacent nucleotide, abstrac-
tion from the C1′ and/or C4′ positions is most likely.13 Ad-
dressing this possibility in oligonucleotides was limited by the
necessity of such a process to give rise to direct strand scission
or alkali-labile lesions in order to be detected by gel electro-
phoresis.7 The absence of direct strand breaks under anaerobic
conditions was sufficient to eliminate C4′ hydrogen atom
abstraction by 1 as a possible pathway, because this process is
known to result in elimination of the adjacent phosphate.12b,25
However, the reliance on gel electrophoresis could allow hydro-
gen atom abstraction from the C1′ position to go undetected
since alkali-labile lesions and direct strand breaks are produced
inefficiently from C1′ radicals under anaerobic conditions when
exogenous hydrogen atom donors are present.16 Using the
detritylated form of dinucleotide 4 as a mechanistic probe, we
anticipated that we could employ the stereoselectivity of the
reduction of 2′-deoxyuridin-1′-yl (7) as a means for detecting
internucleotidyl hydrogen atom abstraction by 1. Hydrogen atom
donation to 7 by hydrogen atom donors (e.g., thiols) occurs
preferentially from the si face to yield â-2′-deoxyuridine as the
major product.16b,26 The preference for this process to yield â-2′-
deoxyuridine varies from 1.5 to 8.3, depending on the hydrogen
atom donor and whether the radical is produced in the
monomeric nucleoside, single-stranded DNA, or double-stranded
DNA. We reasoned that R-2′-deoxyuridine formation resulting
from thiol trapping of 7 would serve as a clear indication for
internucleotidyl hydrogen atom abstraction by 5,6-dihydrothy-
midin-5-yl (1) in this system. R-2′-Deoxyuridine, which could
have been observed in amounts as small as 2% relative to the
â-anomer, was not detected when the crude photolysate of a
sample of detritylated 4 that was irradiated in the presence of
â-mercaptoethanol was subjected to enzymatic digestion.16b
Using the observed stereoselectivity for reduction of monomeric
2′-deoxyuridin-1′-yl (7, â:R ) 1.5) and (2.3-8.0) × 106 M-1
s-1 as the rate constant for reaction of 1 with â-mercaptoethanol,
an upper limit for internucleotidyl hydrogen atom abstraction
is estimated to be between 115 and 440 s-1 (eq 3).18,26,27
- •
studies suggested that hydroperoxyl radical elimination (O2
+ H+) and reduction of 3 by exogenous hydrogen atom donors
would be competing processes for internucleotidyl hydrogen
atom abstraction (Scheme 5).17,30,31 Moreover, independent
investigations of the reactivity of 5,6-dihydrothymidin-5-yl (1)
and 2′-deoxyuridin-1′-yl (7) indicated that 17 should be expected
from internucleotidyl C1′ hydrogen atom abstraction by the
peroxyl radical 3 in the presence of a reducing agent.6,7,16-18,24,26
As expected from studies on monomeric 1, dinucleotides
containing thymidine (14) and thymidine C5 hydrate (15) were
observed under aerobic conditions, the former due at least in
part to hydroperoxyl radical elimination from 3 (Table 2). Most
significant is the observation of 17, the product expected from
the O2-dependent DNA damage amplification process under
reducing conditions (Scheme 1). The selective reduction of the
diastereomeric peroxyl radicals to yield (5S)- and (5R)-15
exhibited a dependence on hydrogen atom donor similar to that
of the formation of 13 under anaerobic conditions (Table 3).
Trapping of the diastereomeric peroxyl radicals by thiol
competes with elimination to 14. Formation of the 5R-
diastereomer of 15 is favored by attack from the less hindered
face of 3 by t-BuSH. In contrast, we believe that hydrogen
bonding of â-mercaptoethanol to the uracil component of the
dinucleotide results in preferential trapping of the diastereomeric
peroxyl radical that yields (5S)-15. Inherent assumptions in this
analysis are that O2, which reacts with 1 at close to diffusion-
controlled rates, produces 3 with little or no facial selectivity,
and that the diastereomeric peroxyl radicals eliminate superoxide
at comparable rates.
Although the identity of the signature product of DNA
damage amplification (17) was corroborated via co-injection
(28) Eibenberger, J.; Schulte-Frohlinde, D.; Steenken, S. J. Phys. Chem.
1980, 84, 704.
(29) (a) Jagannadham, V.; Steenken, S. J. Am. Chem. Soc. 1988, 110,
2188. (b) Steenken, S.; Jagannadham, V. J. Am. Chem. Soc. 1985, 107,
6818.
(30) (a) Chatgilialoglu, C.; Timokhin, V. I.; Zaborovskiy, A. B.; Lutsyk,
D. S.; Prystansky, R. E. J. Chem. Soc., Chem. Commun. 1999, 405. (b)
Chatgilialoglu, C.; Timokhin, V. I.; Zaborovskiy, A. B.; Lutsyk, D. S.;
Prystansky, R. E. J. Chem. Soc., Perkin Trans. 2 2000, 577.
(31) (a) Hildebrand, K.; Schulte-Frohlinde, D. Int. J. Radiat. Biol. 1997,
71, 377. (b) Schulte-Frohlinde, D.; Behrens, G.; O¨ nal, A. Int. J. Radiat.
Biol. 1986, 50, 103. (c) Chenier, J. H. B.; Furimsky, E.; Howard, J. A.
Can. J. Chem. 1974, 52, 3682.
(25) Giese, B.; Dussy, A.; Meggers, E.; Petretta, M.; Schwitter, U. J.
Am. Chem. Soc. 1997, 119, 11130.
(26) Goodman, B. K.; Greenberg, M. M. J. Org. Chem. 1996, 61, 2.
(27) Newcomb, M. Tetrahedron 1993, 49, 1151.

5186 J. Am. Chem. Soc., Vol. 123, No. 22, 2001
Tallman and Greenberg
Table 2. Aerobic Photolysis of Phenyl Selenide Dinucleotide (4)a
mass
yield (%)b
balance
trap
13
14
15
17
(%)b
â-mercapto- 4 ( 2 28 ( 17
ethanol
33 ( 3 0.15 ( 0.10 65 ( 22
t-Butyl thiol 1.5 ( 1.5 33 ( 0.1 10.5 ( 0.5 0.13 ( 0.05 52 ( 1
a [4] ) 0.1 mM; [trap] ) 1.1 mM; solvent, CH3CN/H2O, 1:1 by
volume. b Reported yields and mass balances represent average yields
from individual experiments ( the standard deviation measured from
these averages.
Table 3. Stereoselectivity for the Formation of 15 and 17 from the
Phenyl Selenide Dinucleotidea
trap
(5S)-15:(5R)-15b
(5S)-17:(5R)-17b
â-mercaptoethanol
tert-butyl thiol
2.4 ( 1.0
0.9 ( 0.2
6.4 ( 0.1
c
a [4] ) 0.1 mM; [trap] ) 1.1 mM; solvent, CH3CN/H2O, 1:1 by
volume. b Reported selectivities represent averages from individual
experiments ( the standard deviation measured from these averages.
c The amount of (5R)-17 was below the limit of detection.
Figure 1. Illustration of the greater proximity of the peroxyl radical
in (5S)-3 to the C1′ hydrogen of the adjacent nucleotide compared to
the epimeric peroxyl radical ((5R)-3).
on reverse-phase HPLC with independently synthesized material,
further evidence for its formation was desired due to the small
amount of product formed during the photolysis of 4. This
evidence was obtained by taking advantage of a recent study
in which oligonucleotides containing 2-deoxyribonolactone were
shown to give rise to condensation products with nucleophiles
after undergoing â-elimination to the butenolide.24 In this
previous study, reaction of N,N′-dimethylethylenediamine and
5-O-dimethoxytritylated furanone produced 24. An analogous
butenolide trapping product was detected upon reaction of
2-deoxyribonolactone in an oligonucleotide with this same di-
amine. Treatment of a crude photolysate of 4 with N,N′-dimeth-
ylethylenediamine yielded 24 (confirmed via co-injection with
independently synthesized material) in a yield comparable (0.41
( 0.15%) to that in which 17 was observed. Taken together,
these results confirm that the lactone-containing product 17 is
produced upon generation of 1 under aerobic conditions in the
presence of exogenous reducing agents, and provides strong
evidence for the previously proposed O2-dependent DNA
damage amplification mechanism (Scheme 1).7b,c
Furthermore, given the number of assumptions involved in
arriving at these approximate rate constants, it is quite possible
that the difference between the independent estimates is
insignificant. The small rate constant and corresponding low
yield of product resulting from internucleotidyl hydrogen atom
abstraction caused us to consider the possibility that 17 was
formed via an alternative pathway, such as one involving a
diffusible species. Although precautions were taken to eliminate
metal ions in the reaction, we addressed the possibility that
superoxide released during the formation of 14 was transformed
via Haber-Weiss and Fenton chemistry into hydroxyl radical.
The hydroxyl radical could subsequently abstract the C1′
hydrogen from the 2′-deoxyuridine of a molecule of 15 that
was produced independently from 4. This possibility was
eliminated by the absence of an effect by mannitol (a known
hydroxyl radical scavenger) on the yield of 17.3a Moreover,
further evidence in support of the intramolecular mechanism
was gleaned from photolyses carried out in the presence of a
competing dinucleotide (TpT). The yield of 17 was unchanged
when 4 (0.1 mM) and â-mercaptoethanol (1.1. mM) were
photolyzed in the presence of 0.1 mM (yield of 17 ) 0.15 (
0.10%) or 0.2 mM (yield of 17 ) 0.21 ( 0.10%) TpT.
The observed stereoselectivity for formation of 17 is also
consistent with the mechanism involving internucleotidyl C1′
hydrogen atom abstraction by a selective nucleobase peroxyl
radical (Table 3). Formation of 17 containing the 5S-diastere-
omer of thymidine C5 hydrate is significantly favored over that
of the dinucleotide containing the epimeric modified thymidine.
This is consistent with the expected better proximity of the 5S-
peroxyl radical ((5S)-3) to the C1′ hydrogen atom of the
deoxyuridine bonded to its 5′-phosphate than that of the 5R-
peroxyl radical (Figure 1), and is clearly visible in a three-
dimensional model of detritylated (5S)-3.34 The 5R-peroxyl
The relative yields of thymidine C5 hydrate (15)- and
thymidine (14)-containing dinucleotides compared to lactone
product indicate that the rate constant for internucleotidyl
hydrogen atom abstraction by 3 is modest. A more quantitative
value for the facility of this process was estimated by making
assumptions regarding rate constants for the reactions that
compete with internucleotidyl hydrogen atom abstraction (Scheme
5). The approximate rate constant obtained using the competition
between hydroperoxyl radical elimination to yield 14 (kelim) and
internucleotidyl hydrogen atom abstraction (kabs) to form 17 is
estimated to be kabs ≈ 0.1-0.3 s-1 32
. Estimating the magnitude
of kabs by using the competition with peroxyl radical reduction
(kred) is even less exact, due to the range of rate constants
available for reaction between alkylperoxyl radicals and thiols
((0.4-5.0) × 103 M-1 s-1).31 The measured product ratios ([15]:
[17], Table 3) and literature values for kred suggest as an upper
limit kabs ≈ 0.1 s-1 33
.
These rate constants are approximated
(32) The value estimated for kelim (23.4 s-1) is slightly lower than that
reported previously (see ref 17). This change is made in order to
accommodate the recently measured rate constant for the reaction of the
cumyl peroxyl radical with Bu3SnH ((1.6-1.8) × 103 M-1 s-1; see ref 30)
which is used as an approximation for kred in Scheme 5.
(33) In these calculations, it is assumed that transformation of the initially
trapped peroxyl radical (3) by thiol to 15 and transformation of the radical
formed upon internucleotidyl hydrogen atom abstraction to 17 occur with
equal efficiency.
using only a single concentration of exogenous hydrogen atom
donor (in effect, a single data point) and should be considered
to be crude estimates. Given the low yields of internucleotidyl
hydrogen atom abstraction product 17 and the desire to carry
out experiments under pseudo-first-order reaction conditions in
hydrogen atom donor, it was not deemed practical to carry out
the competition experiments at a variety of thiol concentrations.
(34) See Supporting Information.

Oxygen-Dependent DNA Damage Amplification
J. Am. Chem. Soc., Vol. 123, No. 22, 2001 5187
Upon completion of the reaction, the samples were transferred to
Eppendorf tubes (1.5 mL) and lyophilized. The residue was taken up
in benzophenone (10 µL) and CH3CN/NH4HCO3 buffer (90 µL, 25
mM, pH 6.8, 1:1 by volume). Samples containing tin were extracted
with hexanes (3 × 200 µL). The samples were analyzed by reverse-
phase (C8) HPLC using a gradient of NH4HCO2 (0.2 M, pH 6.2) and
either CH3CN (method A) or MeOH (method B) at a flow rate of 1.0
mL/min. Complete resolution was not obtained under a single gradient
program. Therefore, the samples were analyzed using both methods A
(Supporting Information, Table 1) and B (Supporting Information, Table
2). Typically, 50 µL of the sample was injected using an AUFS of 1.0
and monitored at 240 nm.
The products were quantitated using the calculated response factors
listed below with benzophenone as the internal standard (Supporting
Information, Table 3). The yields of (5R)-15, (5R)-13, (5S)-13, (5R)-
16, (5S)-16, (5R)-17, (5S)-17, and 24 and conversion of 4 were
determined using method A. Under these conditions, (5S)-15 and 14
coelute. Therefore, the samples were also analyzed by method B, in
which (5R)-15 and (5S)-15 coelute but are separable from 14. The yield
of (5S)-15 was obtained by subtracting the yield of (5R)-15 determined
using method A from the total yield found in method B. The yield of
14 was determined by subsequently subtracting (5S)-15 from the total
yield of 14 + (5S)-15 determined in method A.
Anaerobic Photolysis of Detritylated 4 and Subsequent Enzy-
matic Digestion. Samples containing detritylated 4 (0.1 mM) and
t-BuSH (1.1 mM) in CH3CN/H2O (500 µL, 1:1 by volume) were
degassed and photolyzed at 350 nm (96 W) for 1 h. The samples were
transferred to Eppendorf tubes and lyophilized. One set of samples was
taken up in a solution of 2′-deoxycytidine (10 µL) and H2O (90 µL)
and analyzed directly by reverse-phase (C18) HPLC. The second set
was subjected to enzymatic digest. The residue was taken up in buffer
(86 µL, Tris-HCl, pH 7.9, 50 mM; NaCl, 100 mM; DTT, 1 mM) and
MgCl2 (100 mM). Nuclease P1 (10 µL, 0.3 unit/µL), snake venom
phosphodiesterase (2 µL, 0.003 unit/µL), and calf intestine alkaline
phosphatase (2 µL, 10 units/µL) were added to the sample and incubated
at 37 °C for 12 h. 2′-Deoxycytidine (10 µL) was added to the sample
and analyzed by reverse-phase (C18) HPLC using method C (Table
4). The sample was analyzed at 205 nm at an AUFS of 1.0 with dC as
the internal standard. The response factors listed below were estimated
on the basis of the extinction coefficients of dU and dC at 205 nm
(Supporting Information, Table 5).
Anaerobic Photolysis of 4 in the Presence of Tetranitromethane.
Samples containing 4 (0.15 mM), tetranitromethane (1.04 mM), and
âME (1.4 mM) were prepared in CH3CN/NH4HCO2 buffer (25 mM,
pH 6.8, 1:1 by volume) with a final volume of 500 µL. The samples
were degassed using the freeze-pump-thaw method (four 3-min
cycles) and photolyzed at 350 nm (96 W) for 1 h. The samples were
transferred to Eppendorf tubes and lyophilized. The residue was taken
up in benzophenone (10 µL) and CH3CN/NH4HCO2 buffer (90 µL, 25
mM, pH 6.8, 1:1 by volume) and analyzed by reverse-phase (C8) HPLC
using methods A and B (Supporting Information, Tables 1 and 2), as
discussed above.
radical ((5R)-3) is drawn in a pseudoequatorial conformation
(Figure 1), because this arrangement of the dihydropyrimidine
ring brings the peroxyl radical marginally closer to the 5′-
nucleotide. The (5R)-3 radical must adopt the syn conformation
in order to bring the radical center into close proximity of the
5′-adjacent nucleotide’s C1′ hydrogen atom. Although the barrier
to such a conformational change in this molecule is unknown,
UV-melting studies, NMR experiments, and molecular dynamics
simulations on related dihydropyrimidine molecules ((5R)-
thymidine C5 hydrate and 2′-deoxy-5,6-dihydrouridine) in DNA
are consistent with these molecules existing in DNA as the
respective anti isomers.35-38
Conclusions
These experiments provide additional evidence for the
formation of tandem lesions from a nucleobase radical (5,6-
dihydrothymidin-5-yl, 1) that is produced in DNA by γ-radi-
olysis. Independent synthesis of putative products and indepen-
dent generation of 1 in a dinucleotide provides further
substantiation for the requirement of O2 in the amplification of
DNA damage that is initiated by the formation of this radical.
Product analyses suggest that the 5S-peroxyl radical (3) is more
efficient at effecting internucleotidyl hydrogen atom abstraction
than its epimer. This result is not surprising, based upon the
assumptions that the peroxyl radicals exist predominantly in the
anti conformation (Figure 1) and that in the current experiment
there is no 2′-deoxynucleotide bonded to the 3′ position of 1.
The estimated rate constant for the process is modest, and rate
constants are not available for comparable bimolecular reactions
between nucleosides and peroxyl radicals in order to estimate
the effective molarity of the substrate in this system. However,
if one uses the reaction between THF and t-BuOO• as a rough
comparison (kH abs ) 0.085 M-1 s-1), the effective molarity of
•
the abstracted hydrogen atom in 3 is between 1.2 and 3.4 M,
depending upon which estimate of kabs is employed (Scheme
5).39 The rate constant for internucleotidyl hydrogen atom
abstraction (kabs) estimated in this work may be considered a
lower limit for the comparable process in duplex DNA due to
the absence of significant secondary structure in the dinucleotide
substrate. It is possible that internucleotidyl hydrogen atom
abstraction will be faster in duplex DNA, where the more rigid
secondary structure will increase the effective molarity of the
nucleotide bonded to the 5′-phosphate of 1, resulting in higher
yields of lesions analogous to 17. However, it is interesting to
note that the yield of tandem lesion 17 observed in this
structurally minimal system (4) is consistent with those estimated
in plasmid DNA for the formation of similar lesions.40
Aerobic Photolysis of 4 in the Presence of Mannitol. Samples
containing 4 (0.10 mM) and mannitol (1.1 mM) were prepared in
CH3CN/H2O (500 µL, 1:1 by volume) and photolyzed at 350 nm (96
W) for 1 h. The samples were worked up as above and analyzed by
reverse-phase (C8) HPLC using method A (Supporting Information,
Table 1) for formation of 17.
Experimental Section
General Procedure for the Photolysis of 4. Samples containing 4
(0.1 mM) and the appropriate trap (1.1 mM) were prepared in
CH3CN/H2O (1:1 by volume) having a final volume of 500 µL.
Anaerobic samples were degassed using the freeze-pump-thaw
method (four 3-min cycles), and aerobic samples were left open to the
atmosphere. The samples were photolyzed at 350 nm in a Rayonet
photoreactor for the appropriate time.
See Supporting Information for synthetic procedures.
Acknowledgment. Financial support of this work from the
National Institutes of Health (GM-54996) is greatly appreciated.
We thank Brian C. Bales and K. Nolan Carter for technical
assistance, and the reviewers for helpful comments.
(35) Sambandam, A.; Greenberg, M. M. Nucleic Acids Res. 1999, 27,
3597.
(36) Villanueva, J. M.; Pohl, J.; Doetsch, P. W.; Marzilli, L. G. J. Am.
Chem. Soc. 1999, 121, 10652.
(37) Grand, A.; Cadet, J. Acta Crystallogr. 1978, B34, 1524.
(38) Miaskiewicz, K.; Miller, J.; Ornstein, R.; Osman, R. Biopolymers
1995, 35, 113.
(39) Howard, J. A. In The Chemistry of Free Radicals Peroxyl Radicals;
Alfassi, Z. B., Ed.; John Wiley: New York, 1997.
(40) Milligan, J. R.; Aguilera, J. A.; Nguyen, T.-T. D.; Ward, J. F.; Kow,
Y. W.; He, B.; Cunningham, R. P. Radiat. Res. 1999, 151, 334.
Supporting Information Available: Procedures for the
synthesis of all dinucleotides used in this study, and three-
dimensional structures of detritylated 1, (5S)-3, and (5R)-3
(PDF). This material is available free of charge via the Internet
at http://pubs.acs.org.
JA010180S