Model studies of DNA C5′ radicals. Selective generation and reactivity of 2′-deoxyadenosin-5′-yl radical

Article abstract of DOI:10.1021/ja029374d

The reaction of hydrated electrons (e_aq^-) with 8-bromo-2′-deoxyadenosine has been investigated by radiolytic methods coupled with product studies and addressed computationally by means of DFT-B3LYP calculations. Pulse radiolysis revealed that this reaction was complete in ～0.3 μs, and, at this time, no significant absorption was detected. The spectrum of a transient developed in 20 μs has an absorbance in the range 300-500 nm (ε_max? 9600 M^-1 cm^-1 at 360 nm), and it was assigned to aromatic aminyl radical 3. Computed vertical transitions (TD-UB3LYP/6-311+G*) are in good agreement with the experimental observations. Radical 3 is obtained by the following reaction sequence: one-electron reductive cleavage of the C-Br bond that gives the C8 radical, a fast radical translocation from the C8 to C5′ position, and an intramolecular attack of the C5′ radical at the C8,N7 double bond of the adenine moiety. The rate constant for the cyclization is 1.6 × 10⁵ s^-1. On the basis of the theoretical findings, the cyclization step is highly stereospecific. The rate constants for the reactions of C5′ and aminyl 3 radicals with different oxidants were determined by pulse radiolysis methods. The respective rate constants for the reaction of 2′-deoxyadenosin-5′-yl radical with dioxygen, Fe(CN)₆^3-, and MV²⁺ in water at ambient temperature are 1.9 × 10⁹, 4.2 × 10⁹, and 2.2 × 10⁸ M^-1 s^-1. The value for the reaction of aminyl radical 3 with Fe(CN)₆^3- is 8.3 × 10⁸ M^-1 s^-1, whereas the reaction with dioxygen is reversible. Tailored experiments allowed the reaction mechanism to be defined in some detail. A synthetically useful radical cascade process has also been developed that allows in a one-pot procedure the conversion of 8-bromo-2′-deoxyadenosine to 5′,8-cyclo-2′-deoxyadenosine in a diastereoisomeric ratio (5′R):(5′S) = 6:1 and in high yield, by reaction with hydrated electrons in the presence of K₄Fe(CN)₆.

Full text of DOI:10.1021/ja029374d

Published on Web 03/12/2003
Model Studies of DNA C5′ Radicals. Selective Generation and
Reactivity of 2′-Deoxyadenosin-5′-yl Radical
Chryssostomos Chatgilialoglu,* Maurizio Guerra, and Quinto G. Mulazzani
Contribution from the ISOF, Consiglio Nazionale delle Ricerche, Via P. Gobetti 101,
40129 Bologna, Italy
Received November 17, 2002; E-mail: chrys@isof.cnr.it
Abstract: The reaction of hydrated electrons (eaq-) with 8-bromo-2′-deoxyadenosine has been investigated
by radiolytic methods coupled with product studies and addressed computationally by means of DFT-
B3LYP calculations. Pulse radiolysis revealed that this reaction was complete in ∼0.3 µs, and, at this time,
no significant absorption was detected. The spectrum of a transient developed in 20 µs has an absorbance
in the range 300-500 nm (ꢀmax = 9600 M-1 cm-1 at 360 nm), and it was assigned to aromatic aminyl
radical 3. Computed vertical transitions (TD-UB3LYP/6-311+G*) are in good agreement with the
experimental observations. Radical 3 is obtained by the following reaction sequence: one-electron reductive
cleavage of the C-Br bond that gives the C8 radical, a fast radical translocation from the C8 to C5′ position,
and an intramolecular attack of the C5′ radical at the C8,N7 double bond of the adenine moiety. The rate
constant for the cyclization is 1.6 × 105 s-1. On the basis of the theoretical findings, the cyclization step is
highly stereospecific. The rate constants for the reactions of C5′ and aminyl 3 radicals with different oxidants
were determined by pulse radiolysis methods. The respective rate constants for the reaction of
2′-deoxyadenosin-5′-yl radical with dioxygen, Fe(CN)63-, and MV2+ in water at ambient temperature are
1.9 × 109, 4.2 × 109, and 2.2 × 108 M-1 s-1. The value for the reaction of aminyl radical 3 with Fe(CN)6
3-
is 8.3 × 108 M-1 s-1, whereas the reaction with dioxygen is reversible. Tailored experiments allowed the
reaction mechanism to be defined in some detail. A synthetically useful radical cascade process has also
been developed that allows in a one-pot procedure the conversion of 8-bromo-2′-deoxyadenosine to 5′,8-
cyclo-2′-deoxyadenosine in a diastereoisomeric ratio (5′R):(5′S) ) 6:1 and in high yield, by reaction with
hydrated electrons in the presence of K4Fe(CN)6.
Scheme 1
Introduction
Apart from the usual glycosidic bond in 5′,8-cyclopurine-2′-
deoxynucleosides, there is an additional base-sugar linkage
between the C8 position of purine and the C5′ position of the
2-deoxyribose (cf. Scheme 1). The cyclopurine lesions are
observed among the decomposition products of DNA, when
exposed even to moderate doses of ionizing radiations, and their
percentage levels are comparable to the two major lesions, that
is, 8-oxo-7,8-dihydroguanine and thymine glycol.1-6 These
(1) Dirksen, M.-L.; Brakely, W. F.; Holwitt, E.; Dizdaroglu, M. Int. J. Radiat.
Biol. 1988, 54, 195-204. Dizdaroglu, M.; Dirksen, M.-L.; Jiang, H.;
Robbins, J. H. Biochem. J. 1987, 241, 929-932.
(2) Dizdaroglu, M.; Jaruga, P.; Rodriguez, H. Free Radical Biol. Med. 2001,
30, 774-784.
lesions, when formed in DNA, could have a significant
biological impact on the conformation and function of the double
helix.7-10 When DNA, oligonucleotides, or nucleosides are
exposed to hydroxyl radicals, abstraction of the hydrogen occurs
(3) Jaruga, P.; Birincioglu, M.; Rodriguez, H.; Dizdaroglu, M. Biochemistry
2002, 41, 3703-3711.
(4) In polyadenylic acid, see: Fuciarelli, A. F.; Miller, G. G.; Raleigh, J. A.
Radiat. Res. 1985, 104, 3272-283. Fuciarelli, A. F.; Shum, F. Y.; Raleigh,
J. A. Biochem. Biophys. Res. Commun. 1986, 134, 883-887. Fuciarelli,
A. F.; Shum, F. Y.; Raleigh, J. A. Radiat. Res. 1987, 110, 35-44.
(5) It is worth mentioning that the formation of 5′,8-cyclopurine nucleosides
was first discovered to take place in 5′-AMP upon hydroxyl radical attack.
See: Keck, K. Z. Naturforsch. 1968, 23B, 1034-1043. Raleigh, J. A.;
Kremers, W.; Whitehouse, R. Radiat. Res. 1976, 65, 414-422. Raleigh, J.
A.; Fuciarelli, A. F. Radiat. Res. 1985, 102, 165-176. Haromy, T. P.;
Raleigh, J.; Sundaralingam, M. Biochemistry 1980, 19, 1718-1722.
(6) For some reviews, see: Dizdaroglu, M. Free Radical Biol. Med. 1991, 10,
225-242. Breen, A. P.; Murphy, J. A. Free Radical Biol. Med. 1995, 18,
1033-1077. Burrows, C. J.; Muller, J. G. Chem. ReV. 1998, 98, 1109-
1151. Chatgilialoglu, C.; O’Neill. P. Exp. Gerontol. 2001, 36, 1459-1471.
(7) Kuraoka, I.; Bender, C.; Romieu, A.; Cadet, J.; Wood, R. D.; Lindahl, T.
Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 3832-3837.
(8) Brooks, P. J.; Wise, D. S.; Berry, D. A.; Kosmoski, J. V.; Smerdon, M. J.;
Somers, R. L.; Mackie, H.; Spoonde, A. Y.; Ackerman, E. J.; Coleman,
K.; Tarone, R. E.; Robbins, J. H. J. Biol. Chem. 2000, 275, 22355-22362.
(9) Randerath, K.; Zhou, G. D.; Somers, R. L.; Robbins, J. H.; Brooks, P. J.
J. Biol. Chem. 2001, 276, 36051-36057.
(10) Kuraoka, I.; Robins, P.; Masutani, C.; Hanaoka, F.; Gasparutto, D.; Cadet,
J.; Wood, R. D.; Lindahl, T. J. Biol. Chem. 2001, 276, 49283-49288.
9
10.1021/ja029374d CCC: $25.00 © 2003 American Chemical Society
J. AM. CHEM. SOC. 2003, 125, 3839-3848
3839

A R T I C L E S
Chatgilialoglu et al.
mainly at the C4′ and C5′ positions of the sugar moieties.11 It
has been proposed that the C5′ radical might intramolecularly
attack the C8,N7 double bond of the adenine or guanine moieties
to form 5′,8-cyclopurine nucleotides as the final products
(Scheme 1).1-6,12,13 Cyclopurine deoxynucleosides are formed
in two diastereoisomeric forms depending on the configuration
at the C5′ position, that is, (5′S)- and (5′R)-isomers in Scheme
1. Depending on the substrate and the experimental conditions,
the ratio of the (5′S)- and (5′R)-isomers changes substantially.
γ-Irradiation of an aqueous solution of 2′-deoxyadenosine2 or
2′-deoxyguanosine3 afforded mainly the (5′R)-isomers, suggest-
ing a highly stereoselective radical process. The (5′R)-isomer
was also detected under similar conditions upon irradiation of
tetranucleotide d(ApCpGpT).14 Regarding the effect of DNA
conformation, (5′R):(5′S) ratios of approximately 2 and 0.3 were
recently reported by Dizdaroglu and co-workers for cyclo-
deoxyadenosine and cyclodeoxyguanosine, respectively, in
double-stranded DNA,2,3 whereas the (5′R)-isomer of both
compounds was found to predominate in single-stranded DNA.1
It is worth also mentioning that the yields of cyclopurine
nucleosides in single-stranded DNA were higher than those in
double-stranded DNA and that the cyclodeoxyadenosines are
more abundant than cyclodeoxyguanosines.1
Cadet and co-workers recently prepared (5′S)-5′,8-cyclo-dAdo
in seven steps starting from N6-benzoyl-dAdo in an overall yield
<10%,15,16 following the strategy developed by Matsuda et al.
for the synthesis of the ribo analogue.17 The preparation of (5′R)-
5′,8-cyclo-dAdo was achieved by two further steps, involving
inversion of configuration at the C5′ position.18 Synthetic
oligonucleotides that contain these modified derivatives at
selected sites were also prepared to investigate the biochemical
and biophysical features of such lesions.8,15,18 Indeed, recent
experiments indicated that neither the (5′S)- nor the (5′R)-isomer
is recognized by human DNA glycosylases active in the base
excision repair pathway.7,8 The role of nucleotide excision repair
was also studied, and both isomers were found to be relatively
poor substrates, the (5′S)-isomer being repaired more efficiently
than the (5′R)-isomer.7,8 The (5′S)-isomer was found to be a
strong block to gene expression in Chinese hamster ovary and
human cells.8 It was also reported that stereospecific differences
between the (5′S)- and (5′R)-isomer residues affect the relative
resistance to exonucleolytic activity and give rise to different
efficiencies of translesion synthesis by human polymerase η.10
Several dinucleotides containing the 5′,8-cyclo-dAdo moiety
have also been identified in mammalian cellular DNA in vivo,
where its level is enhanced by conditions of oxidative stress.9
It was suggested that in contrast to several other types of
Scheme 2
oxidative DNA damage, cyclopurine nucleotides are chemically
stable and would be expected to accumulate at a slow rate over
years in the DNA, in particular for patients affected by the
genetic disease xeroderma pigmentosum, who are unable to
carry out nucleotide excision repair.9,10
In our laboratory, we envisaged that 8-bromopurine nucleo-
sides were prompt to capture electrons and perhaps rapidly lose
the bromine ion to give the corresponding radical in the C8
position,19,20 which in turn could abstract intramolecularly a
hydrogen from the C5′ position to selectively afford the desired
C5′ radical (Scheme 2).21 If our strategy succeeds, we would
have an efficient entry to C5′ radical chemistry, starting from
8-bromopurine nucleosides which are also commercially avail-
able. Our initial attempt using the 8-bromoguanosine showed
that it behaves in a completely different manner.22 8-Bromo-
dAdo then turned out to be successful, and, in a preliminary
communication,23 we reported the UV-visible spectra of the
cyclized aminyl radical and the rate constant for its formation,
together with an extraordinary radical cascade reaction that
afforded the (5′R)-isomer in a one-pot reaction and in excellent
yield.24 We report herein a detailed study which addresses the
chemical reactivities of the 2′-deoxyadenosin-5′-yl radical.
Results and Discussion
Reaction of Hydrated Electrons (eaq-) with 8-Br-dAdo.16
Radiolysis of neutral water leads to the species eaq-, HO•, and
H• as shown in eq 1 where the values in parentheses represent
the radiation chemical yields (G-values) in units of µmol/J.25
-
The reactions of eaq with the substrates were studied by
irradiating deareated solutions containing 0.25 M t-BuOH. With
this amount of t-BuOH, HO• is scavenged efficiently (eq 2, k2
) 6.0 × 108 M-1 s-1), whereas the H• reacts only slowly.26
Therefore, the reactions of H• may be relevant because they
can account for as much as ∼20% of the products.
-
The rate constant for the reaction of eaq with 8-Br-dAdo
was determined by measuring the rate of the optical density
(11) (a) In DNA, abstraction preference for individual hydrogens of the sugar
moieties was found in the order H5′ ≈ H4′ > H2′ ≈ H3′ > H1′, see:
Pogozelski, W. K.; Tullius, T. D. Chem. ReV. 1998, 98, 1089-1108. Sy,
D.; Savoye, C.; Begusova, M.; Michalik, V.; Charlier, M.; Spotheim-
Maurizot, M. Int. J. Radiat. Biol. 1997, 72, 147-155. (b) Computational
data in a simple nucleotide indicate that the most probable sites of H-atom
abstraction by radical species are H5′ and H4′ positions, see: Toure, P.;
Villena, F.; Melikyan, G. G. Org. Lett. 2002, 4, 3989-3992.
(12) Marriagi, N.; Cadet, J.; Teoule, R. Tetrahedron 1976, 32, 2385-2387.
(13) Dizdaroglu, M. Biochem. J. 1986, 238, 247-254.
(14) Schroder, E. E.; Budzinski, J. C.; Wallace, J. D.; Zimbrick, J. D.; Box, H.
C. Int. J. Radiat. Biol. 1995, 68, 509-523.
(19) For a recent review, see: Save´ant, J.-M. In AdVances in Physical Organic
Chemistry; Tidwell, T. T., Ed.; Academic Press: New York, 2000; Vol.
35, pp 117-192.
(20) Such a reaction is well established for bromo derivatives including
5-bromouracil. For example, see: Nese, C.; Yuan, Z.; Schuchmann, M.
N.; von Sonntag, C. Int. J. Radiat. Biol. 1992, 62, 527-541. Sugiyama,
H.; Fujimoto, K.; Saito, I. J. Am. Chem. Soc. 1995, 117, 2945-2946. Cook,
G. P.; Greenberg, M. M. J. Am. Chem. Soc. 1996, 118, 10025-10030.
Fujimoto, K.; Sugiyama, H.; Saito, I. Tetrahedron Lett. 1998, 39, 2137-
2140.
(21) This reaction is expected to be ca. 20 kcal/mol exothermic, see: Berkowitz,
J.; Ellison, G. B.; Gutman, D. J. Phys. Chem. 1994, 98, 2744-2765.
(22) Ioele, M.; Bazzanini, R.; Chatgilialoglu, C.; Mulazzani, Q. G. J. Am. Chem.
Soc. 2000, 122, 1900-1908.
(15) Romieu, A.; Gasparutto, D.; Molko, D.; Cadet, J. J. Org. Chem. 1998, 63,
5245-5249.
(16) dAdo is an abbreviation of 2′-deoxyadenosine, and it will be used throughout
the article.
(23) Flyunt, R.; Bazzanini, R.; Chatgilialoglu, C.; Mulazzani, Q. G. J. Am. Chem.
Soc. 2000, 122, 4225-4226.
(17) Matsuda, A.; Tezuka, M.; Niizuma, K.; Sugiyama, E.; Ueda, T. Tetrahedron
1978, 34, 2633-2637.
(24) It was reported in a note as unpublished results that photolysis at 254 nm
of a deaerated aqueous solution of 8-Br-dAdo afforded (5′S)- and (5′R)-
isomers in 2 and 12% yields, respectively.15
(18) Romieu, A.; Gasparutto, D.; Cadet, J. Chem. Res. Toxicol. 1999, 12, 412-
421.
9
3840 J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003

Model Studies of DNA C5′ Radicals
A R T I C L E S
Scheme 3
-
Figure 1. Plot of kobs for the decay of eaq at 720 nm as a function of
[8-Br-dAdo] in Ar-purged solutions containing 0.25 M t-BuOH at pH ≈ 7.
Inset: Time dependence of absorption at 720 nm in the presence of 0.41
mM 8-Br-dAdo; optical path ) 2.0 cm, dose ) 19.3 Gy. The solid line
represents the first-order kinetic fit to the data.
concentration (0.2-1 mM) and slightly increased with dose/
pulse. This dose dependence was due to the mixing of first-
order growth and second-order decay of the species. By
extrapolation to zero dose, a rate constant k ) (1.6 ( 0.2) ×
105 s-1 was obtained. Inset b shows the effect of the dose on
Figure 2. Absorption spectrum obtained from the pulse radiolysis of an
Ar-purged solution containing 1 mM 8-Br-dAdo and 0.25 M t-BuOH at
pH ≈ 7, taken 17 µs after the pulse; dose ) 18 Gy, optical path ) 2.0 cm.
Insets: (a) Time dependence of absorption at 360 nm; the solid line
represents the first-order kinetic fit to the data. (b) Dependence of ꢀapp (see
text) at 360 nm from the radiation dose.
the apparent molar extinction coefficient at 360 nm (ꢀapp
)
calculated by assuming a radiation chemical yield G ) 0.27
µmol/J, which is the G of hydrated electrons.25 An extinction
coefficient of 9600 ( 200 M-1 cm-1 at 360 nm was calculated
by extrapolating to zero dose. For the second peak, an ꢀ470 nm
of ca. 1500 M-1 cm-1 was calculated. The transient decayed
decrease of eaq at 720 nm (ꢀ ) 1.9 × 104 M-1 cm-1 27
)
as a
-
function of nucleoside concentration. Figure 1 (inset) shows the
disappearance of eaq- in the presence of 0.41 mM of 8-Br-dAdo,
which leads to a pseudo-first-order rate constant, kobs. From the
slope of kobs versus [8-Br-dAdo], the bimolecular rate constant
was found to be (1.6 ( 0.1) × 1010 M-1 s-1 (Figure 1). As
compared with the analogous reaction with dAdo (k ) 8.2 ×
109 M-1 s-1),28 the presence of bromine increased the rate
constant of the reduction by a factor of 2.
by second-order kinetics with k/ꢀ360 nm ) 8.8 × 104 cm s-1 29
.
We assigned the transient shown in Figure 2 to the conjugated
aminyl radical 3 and the observed first-order growth to the
cyclization of radical 2 (Scheme 3), for the following reasons:
-
(i) The fast reaction of eaq with 8-Br-dAdo should afford the
An aqueous solution of 8-Br-dAdo (1 mM) and t-BuOH (0.25
M) at pH ≈ 7 was prepared. Under Ar-purged conditions, the
aryl radical 1 that has the right geometrical arrangement for a
fast exothermic radical translocation to give the C5′ radical 2
(vide infra). (ii) It is well known that neither aryl radical 1 nor
R-hydroxylalkyl radical 2 possesses any significant absorbance
in the region 300-500 nm;31 this was also confirmed by the
reaction of eaq- with 8-bromoadenine (without the sugar moiety).
(iii) The absorption spectrum in Figure 2 strongly resembles in
-
reaction of 8-Br-dAdo with eaq was complete in ca. 300 ns.
At this time, no significant absorption was detected in the 300-
750 nm region. However, a spectrum containing two bands
centered at 360 and 470 nm, respectively, developed in ∼20
µs (Figure 2). The time profile for the formation of the transient
with λmax ) 360 nm (inset a) followed first-order kinetics with
a rate constant (kexpt) that was independent of 8-Br-dAdo
(29) The observed second-order kinetics can be assigned to self-termination of
•
aminyl radicals 3 and/or its cross-termination with the HOC(CH3)2CH2
radical. Taking into consideration the extinction coefficient, we calculated
(25) Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. J. Phys.
Chem. Ref. Data 1988, 17, 513-886 and references therein.
k ) 8.4 × 108 M-1 s-1, which is similar to the reported self-termination
rate constant of Me2C(OH)CH2 radicals (k ) 6.5 × 108 M-1 s-1).30
•
(26) The rate constant of the HO• radical with 8-Br-dAdo was found to be 5.6
× 109 M-1 s-1 (unpublished work). Therefore, more than 96% of HO•
radicals are scavenged by t-BuOH when [8-Br-dAdo] e 1 × 10-3 M.
(27) Hug, G. L. Natl. Stand. Ref. Data Ser., Natl. Bur. Stand. 1981, 69, 6.
(28) Hissung, A.; von Sonntag, C.; Veltwisch, D.; Asmus, K.-D. Int. J. Radiat.
Biol. 1981, 39, 63-71.
(30) Ross, A. B.; Mallard, W. G.; Helman, W. P.; Buxton, G. V.; Huie, R. E.;
Neta, P. NDRL-NIST Solution Kinetic Database - Ver. 3; Notre Dame
Radiation Laboratory, Notre Dame, IN and NIST Standard Reference Data,
Gaithersburg, MD, 1998 and references therein.
(31) Chatgilialoglu, C. In Handbook of Organic Photochemistry; Scaiano, J.
C., Ed.; CRC Press: Boca Raton, FL, 1989; Vol. 2, pp 3-11.
9
J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003 3841

A R T I C L E S
Chatgilialoglu et al.
shape and ꢀ-values that assigned to the isostructural radical 6,
obtained by the reaction of adenine derivatives either with H•
-
or with eaq followed by protonation at C8; in the case of
adenosine, λmax’s at 355 and 440 nm were observed with
extinction coefficients of 10 500 and 1100 M-1 cm-1, respec-
tively.32 The above rationalization was further supported by three
experiments: (i) no kinetic isotopic effect was found for the
formation of transient by replacing H2O with D2O as the reaction
medium, (ii) the same spectrum and the same kinetics were
obtained by replacing 0.25 M t-BuOH with 0.13 M i-PrOH,
which efficiently trap both H• and HO• species;33 therefore, H•
atoms apparently do not react with 8-Br-dAdo, and (iii) no
reaction was observed for H• with 8-Br-dAdo, when the solution
was saturated with N2O to eliminate hydrated electrons. Because
H• reacts with adenosine with a rate constant of 2.6 × 108 M-1
s-1 by adding at the C8 position to give the adduct 6,32 probably
the presence of bromine substituent at this position either
decreases the rate constant substantially or allows the adduct
to release Br• rapidly (vide infra).
3-
Figure 3. Dependence of 1/∆Absorbance at 360 nm on [Fe(CN)6
]
obtained from the pulse radiolysis of Ar-purged solutions
containing 1 mM 8-Br-dAdo and 0.25 M t-BuOH at pH ≈ 7;
optical path ) 2.0 cm, dose ) 23.2 Gy. Inset: Dependence of
∆Absorbance at 360 nm on [Fe(CN)63-]. The solid line
represents an exponential fit to the data.
Redox Properties of the Transient Species. Useful informa-
tion on radicals, including the differentiation between isomeric
forms, can often be obtained by investigating their reactivity
toward so-called redox indicators.34 The reactivity of the
Fe(CN)63-, E°[Fe(CN)63-/Fe(CN)64-] ) 0.36 V,35 was tested
by pulsing deaerated solutions of 1 mM 8-Br-dAdo containing
0.25 M t-BuOH and different concentrations of K3Fe(CN)6 (25-
Figure 4. Dependence of kobs for decay at 360 nm on [Fe(CN)63-] obtained
from the pulse radiolysis of Ar-purged solutions containing 1 mM 8-Br-
dAdo and 0.25 M t-BuOH at pH ≈ 7. Inset: Time dependence at 360 nm
in the presence of 25 mM [Fe(CN)63-]; optical path ) 2.0 cm, dose )
24.4 Gy. The solid line represents the first-order kinetic fit to the data.
-
100 µM) at pH 7. Under these conditions, the eaq reacted
-
exclusively with 8-Br-dAdo, because k[eaq + Fe(CN)63-] )
3.1 × 109 M-1 s-1 25,30
.
The absorption at 360 nm was found to
decrease by increasing the concentration of Fe(CN)63- (see inset
of Figure 3). Using a Stern-Volmer type of approach, we
obtained koxid/kc ) 2.6 × 104 M-1 (Figure 3). The combination
of these data with the kc value yielded koxid ) 4.2 × 109 M-1
4-
Next we considered the reactivity of the reductant Fe(CN)6
.
Ar-purged solutions of 1 mM 8-Br-dAdo containing 0.25 M
t-BuOH and different concentrations of K2Fe(CN)6 (25-100
µM) at pH 7 were used. No substantial changes were observed
with respect to the experiment without the reductant. Indeed,
only a slight decrease (4-5%) of the absorption at 360 nm was
observed in comparison with the analogous reaction of Fe3+
salt (cf. inset of Figure 3). Under these conditions, the kinetics
of formation and decay of transient absorbing at 360 nm were
s-1. Because the reaction between the Me2C(OH)CH2 radical
•
(eq 2) with Fe(CN)6 is slow (k ) 3 × 106 M-1 s-1),36 the
3-
decrease of absorption at 360 nm was attributed to a reaction
3-
of the C5′ radical 2 with Fe(CN)6 (Scheme 3). On the other
hand, the disappearance of the transient at 360 nm followed
3-
first-order kinetics in the presence of Fe(CN)6 (see inset of
unaffected, suggesting slow reactions, if any, of the C5′ radical
Figure 4). From the slope of kobs versus [Fe(CN)63-], the
bimolecular rate constant was found to be (8.3 ( 0.3) × 108
M-1 s-1 (Figure 4). Scheme 3 summarizes our finding; that is,
4-
2 and of the aminyl radical 3 with Fe(CN)6
.
The reactivity of the weaker oxidant methyl viologen (MV2+),
E°(MV2+/MV•+) ) -0.45 V,37 with the transient species was
also investigated. The experiments were performed by adding
50 or 100 µM MV2+ to Ar-purged solutions containing 1 mM
8-Br-dAdo and 0.25 M t-BuOH. Under these conditions, the
3-
Fe(CN)6 was able to oxidize both the C5′ radical 2 and the
aromatic aminyl radical 3. The resulting cations 4 and 5 could
be expected to deprotonate rapidly, affording the corresponding
aldehyde and 5′,8-cyclo-dAdo.
competition between 8-Br-Ado and MV2+ for eaq was com-
-
-
parable because k(eaq + MV2+) ) 7.2 × 1010 M-1 s-1 25,30
.
(32) Candeias, L. P.; Steenken, S. J. Phys. Chem. 1992, 96, 937-944.
(33) k[HO• + i-PrOH] ) 1.9 × 109 M-1 s-1 and k[H• + i-PrOH] ) 7.4 × 109
Indeed, a two-component increase of the absorption at 605 nm,
characteristic of the methyl viologen radical cation (MV+•, ꢀ605
nm ) 1.37 × 104 M-1 cm-1),38 was observed upon irradiation.
In particular, using 100 µM MV2+, we found that the instan-
M-1 s-1 25,30
.
(34) Buxton, G. V.; Mulazzani, Q. G. In Electron Transfer in Chemistry; Balzani,
V., Ed.; Wiley-VCH: Weinheim, 2001; Vol. 1, p 549. For applications to
analogous radicals, see refs 28, 32, and 36.
(35) Hausler, K. E.; Lorenz, W. J. In Standard Potentials in Aqueous Solution;
Bard, A. J., Parsons, R., Jordan, J., Eds.; Marcel Dekker: New York, 1985;
p 408.
(36) Candeias, L. P.; Wolf, P.; O’Neill, P.; Steenken, S. J. Phys. Chem. 1992,
96, 10302-10307.
(37) Wardman, P. J. Phys. Chem. Ref. Data 1989, 18, 1637-1755 and references
therein.
(38) Watanabe, T.; Honda, K. J. Phys. Chem. 1982, 86, 2617-2619.
9
3842 J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003

Model Studies of DNA C5′ Radicals
A R T I C L E S
Figure 6. Dependence of kobs for second-order decay at 360 nm obtained
from the pulse radiolysis of solutions containing 1 mM 8-Br-dAdo and 0.25
M t-BuOH at pH ≈ 7 saturated with Ar/O2 mixtures. Inset: Time
dependence of ∆Absorbance at 360 nm in the presence of 75 mM O2; optical
path ) 2.0 cm, dose ) 17 Gy. The solid line represents a second-order
kinetic fit to the data.
Figure 5. Dependence of ∆Absorbance at 360 nm on [O2] obtained from
the pulse radiolysis of solutions containing 1 mM 8-Br-dAdo and 0.25 M
t-BuOH at pH ≈ 7 saturated with Ar/O2 mixtures; optical path ) 2.0 cm,
dose ) 17 Gy. The solid line represents an exponential fit to the data.
Insets: (a) Dependence of 1/∆Absorbance at 360 nm on [O2]. (b)
Dependence of kobs for growth at 360 nm on [O2].
Table 1. Rate Constants for the Reaction of the
2′-Deoxyadenosin-5′-yl Radical and the 1,2-Dihydroxyethyl
Radicala
taneous formation of MV+• accounted for 32% of eaq-, whereas
the slow growth accounted for 12%. Because the Me2C(OH)-
rate constant (M-1 s-1
)
•
CH2 radical does not react with MV2+, this demonstrates an
b
trapping agent
2′-deoxyadenosin-5′-yl
HOCH(•)CH OH
2
electron transfer between the precursor of aminyl radical 3 (i.e.,
C5′ radical) and MV2+ with a rate constant of 2.2 × 108 M-1
s-1. It is also worth mentioning that under these conditions the
MV•+ decays by second-order kinetics (kobs ) 2.4 × 109 M-1
s-1), which probably involves an electron going back to the
oxygen
(1.8 ( 1) × 109
(4.2 ( 0.4) × 109
(2.2 ( 0.4) × 108
3.2 × 109
3.6 × 109
1.8 × 108
3-
Fe(CN)6
MV2+
a Reactions in neutral water at (22 ( 2) °C. b Data from ref 40; see also
ref 30.
•
aminyl 3 or to the Me2C(OH)CH2 radical.
Table 2. Rate Constants for the Reactions of Aminyl Radicals 3,
6, and 7a
The reactivity of molecular oxygen was investigated by
pulsing solutions of 1 mM 8-Br-dAdo containing 0.25 M
t-BuOH and different concentrations of O2 (0-250 µM) at pH
7. The absorption at 360 nm was found to decrease by increasing
the concentration of O2 (Figure 5). For this decrease, the
rate constant (M-1 s-1
)
trapping agent
radical 3
radical 6d
radical 7e
oxygen
b
3.3 × 108
5.9 × 108
<107
3-
Fe(CN)6
(8.3 ( 0.3) × 108
reactions of eaq- with 8-Br-dAdo and O2 should be considered,
MV2+
c
<107
-
because k[eaq + O2] ) 1.9 × 1010 M-1 s-1 25,30
.
Taking into
-
consideration the percentage of eaq captured by O2 and using
a Reactions in neutral water at (22 ( 2) °C. b See text. c Reaction not
observed. d Data from ref 32. e Data from ref 36.
a Stern-Volmer type of approach (inset a), we obtained koxygen
/
kc ) 6.9 × 103 M-1. A combination of these data with the kc
value yielded koxygen ) 1.1 × 109 M-1 s-1. On the other hand,
the first-order growth (kobs) increased linearly by increasing the
oxygen concentration, that is, kobs ) kc + koxygen[O2]. From the
slope of kobs versus [O2], the bimolecular rate constant was found
to be 2.5 × 109 M-1 s-1 (inset b). An average value of koxygen
) (1.8 ( 1) × 109 M-1 s-1 is suggested for the reaction of the
C5′ radical 2 with molecular oxygen.
Table 1 summarizes the results obtained for the trapping of
2′-deoxyadenosin-5′-yl radical (2), together with the analogous
reactions of the HOCH(•)CH2OH radical produced from eth-
ylene glycol. The reactivities of the two species toward oxygen
and oxidants are very similar. The rate constants for the reactions
with oxygen are typical for R-heteroatom substituted alkyl
radicals.30,40 The rate constants for the reactions with MV2+ are
at least 1 order of magnitude slower than those with the stronger
oxidant Fe(CN)63-, which probably involve an inner-sphere
electron-transfer process.41
The presence of O2 also influenced the disappearance of
radical 3. The second-order kinetics became faster as the
concentration of O2 increased. Figure 6 shows the dependence
of kobs on [O2], and the inset shows the second-order kinetic fit
to the data for 75 µM of oxygen. Such a behavior is consistent
with the reversible reaction of aminyl radical 3 with O2 to give
the corresponding peroxyl adduct (eq 3/-3), followed by cross-
termination between this radical with a second radical 3 (eq 4).
Table 2 summarizes the results obtained for the trapping of
aminyl radical 3. For comparison, the analogous values for the
reactions of isostructural radicals 6 and 7 are also given. The
reactivity of the most structurally similar radical 6 is unfortu-
nately limited to the reaction with MV2+. Indeed, MV2+ is not
Under these conditions, the second-order rate constant, k(so)obs
,
(39) For kinetic details of a similar mechanistic scheme, see: Kelly, C. A.; Blinn,
E. L.; Camaioni, N.; D’Angelantonio, M.; Mulazzani, Q. G. Inorg. Chem.
1999, 38, 1579-1584 and 2756.
will be faster and a complex function of [O2], k3, k-3, and k4.39
(40) Adams, G. E.; Willson, R. L. Trans. Faraday Soc. 1969, 65, 2981-2987.
Steenken, S.; Davies, M. J.; Gilbert, B. C. J. Chem. Soc., Perkin Trans. 2
1986, 1003-1010.
3• + O2 h 3-OO•
(3/-3)
(41) Grodkowski, J.; Neta, P.; Schlesener, C. J.; Kochi, J. K. J. Phys. Chem.
1985, 89, 4373-4378.
3-OO• + 3• f products
(4)
9
J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003 3843

A R T I C L E S
Chatgilialoglu et al.
Figure 8. SOMO (A) and LUMO (B) of the electron adduct 8-Br-dAdo-•
computed at the geometry of 8-Br-dAdo at the UB3LYP/6-31G* level.
experimental evidence. Indeed, electron distribution in the
SOMO of the radical anion computed at the geometry of the
neutral molecule (Figure 8A) shows that the unpaired electron
is mainly localized in the five-membered ring at C8, that is, at
the carbon linked to the Br atom. The lowest unoccupied MO
(LUMO) of the anion lies only 1.5 eV at higher energy than
the SOMO and is localized at the C8-Br bond being antibond-
ing in character (Figure 8B). Thus, the unpaired electron tends
to occupy the antibonding σ*C8-Br MO upon relaxing the
structural parameters, favoring the loss of Br-.
Figure 7. Relative stabilities of syn and anti conformers (E, kcal/mol)
together with the outer and inner rotational barriers about the N-glycosidic
bonds (Ea, kcal/mol) at the UB3LYP/6-31G* level. Values corrected for
the difference in ZPVE (cf. ref 42).
able to oxidize fast enough any of these radicals.32,36 On the
other hand, it is gratifying to see the similarity of the behavior
3-
of radicals 3 and 7 toward the stronger oxidant Fe(CN)6
.
Because E°[Fe(CN)63-/Fe(CN)64-] ) 0.36 V,35 the E°(5+/3•)
should be e0.36 V. Moreover, the rate constant observed for
the decay of MV+• might be attributed to the reaction with the
DFT calculations were then performed to determine the
rotational barriers about the N-glycosidic bond in both the 8-Br-
dAdo (8,9) and the C8 radical of σ-type (10,11). Figure 7 shows
the relative stabilizations of the syn and anti conformers as well
as the rotational barriers for the two available paths (outer and
inner) of the interconversion between the two conformers. The
inner refers to the rotation in which the Br moiety or the radical
center passes above the sugar ring and the outer to the opposite
direction. In both systems, the syn conformer is more stable
than the anti of 1.2-1.4 kcal/mol. The comparison of the two
rotational paths for the conversion of 8 to 9 shows that the inner
rotation is more crowded and, therefore, requires a higher barrier.
After the removal of Br atom, the inner rotation barrier for the
conversion of 10 to 11 considerably decreases due to the lower
steric demand. Therefore, the conformer 11 is substantially
populated.
aminyl radical 3. MV+• is a stronger reductant than Fe(CN)6
4-
and perhaps is able to reduce the aminyl radical 3. Because
E°(MV2+/MV•+) ) -0.45 V,37 the E°(3•/3-) should be g-0.45
V.
MO Calculations. MO calculations with the DFT method
were performed to determine the factors that affect the activation
energies of the suggested reactions.42,43 Initially, B3LYP/6-31G*
calculations were carried out on 8-Br-dAdo in the syn (8) and
anti (9) conformation (Figure 7). The syn conformer is computed
to be more stable than the anti as was found experimentally.
This finding was ascribed to the presence of the bulky Br atom.
The vertical electron affinities (VEA) of the more stable syn
conformer (8) are estimated to be negative (-0.9 eV) at the
B3LYP/6-31G* level so that the radical anion could be unstable
in the gas phase.44 Interestingly, when the structural parameters
of the radical anion were optimized, the radical anion of 8 lost
Br-, producing the neutral σ-type radical 10 in accordance with
Next we considered the radical translocation from the C8 to
the position C5′. In the transition state for such a hydrogen
abstraction, the C8‚‚‚H‚‚‚C5′ have to be collinear or nearly so.
The two conformers around the C4′-C5′ bond (12 and 13),
which differ in stability by 1.2 kcal/mol, have this arrangement
(Figure 9). The radical translocations of both conformers are
exothermic of ca. 20 kcal/mol as expected on the thermochemi-
cal grounds21 and occur with a reaction barrier of ca. 3 kcal/
mol to give the C5′ radical of π-type, which is nearly planar
with a very low interconversion barrier. The conformers 14 and
15 of the C5′ radical have essentially the same stability.
(42) DFT calculations were carried out with the Gaussian 98 system of
programs43 running on DEC-Alpha 500 computers employing the B3LYP
functional and the valence double-ú basis set supplemented with polarization
d-function on heavy atoms (B3LYP/6-31G*). The unrestricted wave
function was used for radical species (UB3LYP/6-31G*). The nature of
the transition states was verified by frequency calculations (one imaginary
frequency), and energy values were corrected for the zero-point vibrational
energy (ZPVE).
(43) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M.
A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.; Stratmann,
R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin,
K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi,
R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.;
Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.;
Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz,
J. V.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.;
Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng,
C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.; Gill, P. M. W.;
Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.; Gonzalez, C.; Head-
Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian 98, revision A.7;
Gaussian, Inc.: Pittsburgh, PA, 1998.
Chart 1 shows the four possible aminyl radicals, which can
be obtained for the cyclization of radicals 14 and 15. Cyclization
of radical 14 is expected to produce radicals 16 and 17, that is,
(5′R)-isomers, which have a defined stereochemistry also at
position C8 derived from the back or front attack, respectively.
Analogously, cyclization of radical 15 is expected to produce
radicals 18 and 19, that is, (5′S)-isomers, derived again from
the back or front attack, respectively. The UB3LYP/6-31G*
calculations have been extended to the cyclization of radical
(44) Diffuse functions were not added to the basis set to better describe the
radical anion because in this case the calculations could provide unreliable
results for unstable anions, see: Guerra, M. Chem. Phys. Lett. 1990, 167,
315-319. Guerra, M. J. Phys. Chem. A 1999, 103, 5983-5988.
9
3844 J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003

Model Studies of DNA C5′ Radicals
A R T I C L E S
Figure 10. Geometries of the chair transition state (A) and product radical
(B) for the back attack of radical 14 at the UB3LYP/6-31G* level.
Figure 9. Relative stabilities (E, kcal/mol) of the two σ-type radical
conformers (12,13) and of the two π-type radical conformers (14,15)
together with the activation energies for the radical translocation (Ea, kcal/
mol) at the UB3LYP/6-31G* level. Values are corrected for the difference
in ZPVE.
Chart 1
Figure 11. Geometries of the boat transition state (A) and product radical
(B) for the front attack of radical 14 at the UB3LYP/6-31G* level.
exothermic as the corresponding Hr values of the boat forming
radicals 17 and 19. The angle of attack φTS in the transition
states of 14 and 15 is close to the normal (109.47°), indicating
that the dispositions of the three interactive atoms, that is, C5′-
C8-N7, in all cyclizations are close to the stereoelectronic
requirements according to the Baldwin-Beckwith set of empiri-
cal rules.45 On the other hand, the barriers of transition states
in the chair arrangements (14 f 16 and 15 f 18) are 3-5
kcal/mol lower than those of the corresponding boat due to the
hydrogen bondings with the oxygen in the sugar ring (2.18 and
2.70 Å, respectively) (Figures 10A and 11A). Therefore, the
reaction enthalpies and the hydrogen bondings of HO5′ in the
chair transition states play an important role in determining the
small size of the reaction barrier in the cyclization 14 f 16.
Figure 12 summarizes the progress of reaction from radical
12 to radical 16 via the intermediacy of the C5′ radical 14 (solid
line). That is, the barriers of 3.2 and 8.4 kcal/mol are needed
for the radical translocation and the cyclization steps, respec-
tively. The exothermicity of this stepwise reaction (12 f 14 f
16) is 29.8 kcal/mol. However, the possibility of a “concerted”
mechanism can also be considered. After the TS, during the
path downhill to 14 there could be a point where the two
moieties (i.e., purine and C5′ radical) are very close to the
optimum geometry of the transition state for cyclization. Instead
of relaxing from the crest to give radical 14, they could
undertake the uphill with much less barrier. Calculations showed
that if the cyclization is forced at this stage assuming the C5′-
C8 as reaction coordinate (see 20 in Figure 12), the reaction
Table 3. Reaction Barriers (Ea), Reaction Enthalpies (Hr), Partial
Bond Lengths (dTS), and Angle of Attacks (φTS) in the Transition
States for the Cyclizations of Radicals 14 and 15 Computed at the
UB3LYP/6-31G* Levela
ring
E ,
a
H ,
r
φTS,
cyclization
formation
kcal/mol
kcal/mol
dTS, Å
deg
14 f 16
14 f 17
15 f 18
15 f 19
chair
boat
chair
boat
8.4
13.3
13.5
16.3
-9.3
-4.9
-11.0
-6.9
2.229
2.181
2.238
2.202
108.6
113.3
110.4
110.2
a Ea and Hr values corrected for the difference in ZPVE (cf. ref 42).
C5′. Table 3 shows the values of reaction barriers (Ea), reaction
enthalpies (Hr), partial bond lengths (dTS), and angle of attacks
(φTS) in the transition states. Figures 10A and 11A show the
transition states of radical 14 for the back or front attacks,
respectively. Figures 10B and 11B show the corresponding
product radicals 16 and 17. The back attack induced the chair
conformation in the forming ring (Figure 10), whereas in the
front attack the ring adopts a boat conformation (Figure 11).
The reaction enthalpies (Hr) for the formation of aminyl radicals
in the chair conformation (16 and 18) are almost twice as
(45) Baldwin, J. E. J. Chem. Soc., Chem. Commun. 1976, 734-736. Beckwith,
A. L. J.; Easton, C. J.; Serelis, A. K. J. Chem. Soc., Chem. Commun. 1980,
482-483.
9
J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003 3845

A R T I C L E S
Chatgilialoglu et al.
computed to occur at higher wavelengths (440 nm) and should
be responsible for the weak absorption observed in the visible
region during pulse radiolysis at 470 nm.
Product Studies. A deareated aqueous solution (1 L)
containing 1.5 mM of 8-Br-dAdo (495 mg), 0.25 M t-BuOH,
and 4 mM K4Fe(CN)6 at pH ≈ 7 was γ-irradiated with a total
dose up to 3 kGy at a dose rate of 18 Gy/min. The crude reaction
mixture was passed through ion-exchange resin (Amberlite IRA-
400) to eliminate the iron salts. The HPLC analysis of the
reaction mixture showed that the starting bromide was consumed
55% and that several products were formed, one of which
predominates.49 Workup of the reaction mixture and separation
by RP silica gel chromatography afforded the major product
(5′R)-5′,8-cyclo-dAdo (21) in a 61% yield (based on the
recovered starting bromide). Some of the minor products could
be isolated as pure materials. The 5′,8-cyclo-2′,5′-dideoxyAdo
(23) and (5′S)-5′,8-cyclo-dAdo were spectroscopically character-
ized and together with adenine and dAdo were used for
calibration of the quantitative HPLC analysis (see below).
The isolation of (5′R)- and (5′S)-isomers was previously
obtained by semipreparative HPLC, and a not very well-resolved
1H NMR (200 MHz) was reported in D2O.15,24 In the Experi-
mental Section, a detailed NMR characterization is reported for
both compounds. In particular, proton NMR (400 MHz) spectra
in both D2O and DMSO are compared. In D2O, the chemical
shifts are shifted 0.2-0.4 ppm upfield as compared to DMSO
and ca. 0.1 downfield as compared to the data reported in the
literature. It is worth mentioning that the J4′5′ is zero (in DMSO)
or 1.2 Hz (in D2O) for the (5′R)-isomer and 6.0-6.1 Hz for the
(5′S)-isomer in both solvents. Structural information of (5′R)-
isomer 21 was also obtained from multiple fragmentation
experiments using ESI-Ion Trap MSn. The MS1 spectrum
consisted of the protonated molecular ion (MH+) at m/z 250,
whereas the MS2 spectrum showed three daughter ions. The
two ions at m/z 232 and 214 should be related to the loss of
one and two water molecules from the parent ion, respectively.
The third one at m/z 164 can be assigned to the base fragment
resulting from the cleavages of the glycosidic and C4′-C5′
bonds.50 Indeed, the MS3 spectrum of this ion leads to m/z 136
(protonated adenine) by the loss of a CO group.
Figure 12. Progress or reaction in a stepwise (12 f 14 f 16) or concerted
(12 f 20 f 16) mechanism.
Table 4. Vertical Optical Transitions (Wavelengths λ, Oscillator
Strengths f) for Aminyl Radical 16 Computed at the UB3LYP/
6-311+G*//UB3LYP/6-31G* Level with the TD-DFT Method along
with Experimental λmax Valuesa
λ, nm
f
transition {spin}
exp, nm
440
350
320
300
0.017
0.093
0.068
0.052
pπ(out-of-phase N1,N3) f SOMO(N7) {â}
SOMO(N7) f π*(N1-C2) {R}
lone pair NH2 f SOMO(N7) {â}
pπ(N9) f SOMO(N7) {â}
470
360
a Only transitions with f greater than 0.005 are reported.
barrier drops from 8.4 to 2.3 kcal/mol (dashed line). However,
20 is not a true minimum of the hypersurface. That is, relaxing
the C5′-C8 constrain 20 gives radical 14. Similarly, Ea values
of 13.3, 13.5, and 16.5 kcal/mol in Table 3 drop to 5.2, 6.3,
and 6.8 kcal/mol, respectively.
The optical spectrum (transition wavelength λ and oscillator
strength f) of the aminyl radical 16 was computed with the time-
dependent DFT method which was shown to reproduce the
UV-vis spectra of radicals fairly well.46 Table 4 shows that
the theoretical results (TD-UB3LYP/6-311+G*//UB3LYP/6-
31G*) are in good accordance with experiment, the transition
energy being overestimated by about 0.15 eV. The higher
intensity transition was computed at 350 nm, that is, very close
to the strong maximum observed at 360 nm in the pulse
radiolysis experiment. Calculations indicate that this transition
is due to an excitation from the SOMO localized to the pπ
orbital of N7 to the antibonding π* of the N1-C2 double bond
(R spin transition). Calculations also suggest that this broad band
that extends to λ < 300 nm could probably originate from the
overlapping of three transitions. The other two transitions are
computed to occur at lower wavelengths (320 and 300 nm) and
involve electron transfer from the N lone pair of the amino group
and from the pπ orbital of the nitrogen (N9) of the glycosyl
link to the SOMO (â spin transition). Transition from the out-
of-phase mixing of the pπ N1 and N3 orbitals to the SOMO is
Additional experiments were performed to obtain further
mechanistic information for this synthetically useful radical
cascade process. The dose profile of the reaction is followed
by HPLC with 3 mL scale experiments, and the products were
quantified by comparison with the authentic samples. For
example, the experiment with 1.5 kGy dose gave the following
results (yields): adenine (5%), (5′R)-isomer 21 (60%), (5′S)-
isomer (10%), dAdo (7%), and compound 23 (7%). Yields are
based on the consumption of 8-Br-dAdo. It is satisfying to see
(47) 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, 9525-9533.
(48) Weisman, J. L.; Head-Gordon, M. J. Am. Chem. Soc. 2001, 123, 11686-
11694.
(49) A detailed dose profile of 8-Br-dAdo consumption and (5′R)-isomer
formation obtained by HPLC analysis showed that the total dose of 3kGy
is the optimum for the product isolation (i.e., high conversion and less
secondary products). The fact that ca. 55% of the starting bromide is
consumed under these conditions indicates that ca. 85% of the produced
solvated electrons and hydrogen atoms react with 8-Br-dAdo.
(50) The ESI-mass spectrum of (5′R)-isomer 21 was previously obtained from
the analysis of the reaction mixture of the HO• radical with dAdo using a
quadrupole and was discussed in detail.2 The fragmentation pattern consisted
of an ion at m/z 164 which has been unequivocally assigned to this
fragmentation.
(46) Good results were previously obtained using both triple ú basis sets47 (TD-
UB3LYP/6-311G**//UHF/6-31G*) and double ú basis sets plus diffuse
functions (TD-UBLYP/6-31+G*//UBLYP/6-31G*).48 The values reported
in Table 2 have been computed with the more flexible 6-311+G* basis set
(TD-UB3LYP/6-311+G*//UB3LYP/6-31G*). Interestingly, the transition
wavelengths differ less than 0.05 nm from those computed with smaller
basis sets (TD-UB3LYP/6-311G**//UB3LYP/6-31G*, TD-UB3LYP/6-
31+G*//UB3LYP/6-31G*).
9
3846 J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003

Model Studies of DNA C5′ Radicals
A R T I C L E S
Scheme 4
Scheme 5
Scheme 5 shows some possible mechanistic paths for the
formation of dAdo by rationalization of the experimental data.
The hydrogen atom might add to the C8,N7 double bond of
the starting bromide to give the radical 24. This radical could
either lose rapidly Br• to afford dAdo (path A) or cleave
heterolytically to give the radical cation of 2′-deoxyadenosine
and Br- (path B). In their turn, Br• or dAdo•+ can react with
4-
Fe(CN)6 to give Br- or dAdo. Alternatively, an adduct that
derived from the reaction of H• with Fe(CN)6 (see below)
might be able to react with 8-Br-dAdo to give dAdo by unknown
chemistry. The mechanism of the adenine formation is unclear
at present.
The role of K4Fe(CN)6 is discussed next. In the presence of
0.25 M t-BuOH and 4 mM K4Fe(CN)6, the attack of HO•
radicals occurs in about 80% of the cases at the t-BuOH (eq 2)
and in about 20% of the cases at the K4Fe(CN)6, k[HO• +
4-
that the ratio (5′R):(5′S) ) 6:1 is identical to that reported in
the photolysis of 8-Br-dAdo.24
The disappearance of the starting material (mol kg-1) divided
by the absorbed dose (1 Gy ) 1 J kg-1) gives the radiation
chemical yield or G[-(8-Br-dAdo)]. A plot of G[-(8-Br-dAdo)]
versus dose and extrapolation of the line to zero dose gives a
G[-(8-Br-dAdo)] ) 0.32 which is close to the sum G(eaq-) +
G(H•) ) 0.27 + 0.06 ) 0.33. These findings lead to the
conclusion that the solvated electrons and the hydrogen atoms
react almost quantitatively with 8-Br-dAdo. The analogous
reaction without K4Fe(CN)6 afforded 21 in 25% yield. Under
these conditions, it was possible to identify and quantify Br-
by comparison with KBr (the HPLC analysis monitored at 210
nm).23 It is gratifying to see that the radiation chemical yield
for the formation of bromide was equal to the consumption of
starting compound. With regards to the latter experiment, when
K4Fe(CN)6 was added to the reaction mixture after its irradiation,
and after a standing time equal to the irradiation time, no
changes were detected with respect to the product yields. The
absence of the post-irradiation effect means that the iron salts
must be involved in the free radical transformation. Furthermore,
the formation of K3Fe(CN)6, which can be conveniently
monitored by UV-vis spectroscopy at 420 nm (ꢀ ) 1020 M-1
cm-1),51 in steady-state experiments was found to be negligible.
Taken into consideration the pulse radiolysis and computa-
tional data for the fate of eaq-, the cyclization of the C5′ radical
should afford mainly the aminyl radical 16 with the defined
stereochemistry at the C5′ and C8 positions (Scheme 4). On
Fe(CN)64-] ) 1.0 × 1010 M-1 s-1 25,30
.
Therefore, the reaction
of HO• with Fe(CN)6 will contribute to the formation of
oxidant Fe(CN)63-. It is also reported that Fe(CN)64- reacts with
H• with a rate constant of 3.9 × 107 M-1 s-1 to give an unknown
adduct.53 The latter could be a reactive species toward the
observed transients or might transform to Fe(CN)63-. Probably
the H• atoms are partitioned between the two reaction channels,
that is, addition to 8-Br-dAdo and to Fe(CN)64-. It is worth
4-
3-
also mentioning that Fe(CN)6 reacts with H• to regenerate
Fe(CN)6 (k ) 6.0 × 109 M-1 s-1) and the carbon-centered
4-
radical generated from eq 2 is also reported to react with
3-
Fe(CN)6 (k ) 3.0 × 106 M-1 s-1 25,30
)
and probably affords
4- 36
Fe(CN)6
.
In summary, a steady-state concentration at the
micromolar level of oxidant(s) should be present during the
course of these reactions.
Conclusions
DNA radicals have attracted the interest of chemists and
biochemists for years, and there is a large body of results for
the involvement of C5′ radicals.6,11,54 In this paper, we studied
the chemistry of model C5′ DNA radicals and in particular of
the 2′-deoxyadenin-5′-yl radical with the aim of determining
the structural and reactivity effects regarding the formation of
the 5′,8-cyclonucleoside as well as the redox properties of the
two participating radicals in this cyclization. We found that the
C5′ radical adds intramolecularly to the C8,N7 double bond of
the adenine moiety with a rate constant of 1.6 × 105 s-1 at
room temperature. This process is highly stereospecific, and
UB3LYP/G-31G* calculations suggest that from the four
possible paths, the one with a chair conformation and R-
configuration at the C5′ position in the forming ring has the
3-
the basis of the pulse radiolysis results, Fe(CN)6 generated
in situ (see below) can easily oxidize radical 16 to give
compound 21. Analogously, radical 17 can afford the same
compound 21, whereas the radical 18 should give the corre-
sponding (5′S)-isomer (cf. Chart 1 and Table 3).52 The radical
16 may be also reduced slowly by Fe(CN)64- followed by fast
protonation to give compound 22, which can readily eliminate
H2O (the OH group in the 5′ position being in anti arrangement
with the H atom in the 8 position) and give the cyclonucleoside
23 after tautomerization (Scheme 4).
(51) Schuler, R. H.; Hartzell, A. L.; Behar, B. J. Phys. Chem. 1981, 85, 192-
199.
(52) The relative yields of (5′R)- and (5′S)-isomers suggest that the gap between
the activation energies for the cyclization of radicals 14 and 15 via the
chair conformation must be much smaller than the difference of calculated
energy barriers (cf. Table 3).
(53) Zehavi, D.; Rabani, J. J. Phys. Chem. 1974, 78, 1368-1373.
(54) For example, see: Goldberg, I. H. Acc. Chem. Res. 1991, 24, 191-198.
Nicolaou, K. C.; Dai, W.-M. Angew. Chem., Int. Ed. Engl. 1991, 30, 1387-
1416. Pratviel, G.; Bernadou, J.; Meunier, B. Angew. Chem., Int. Ed. Engl.
1995, 34, 746-769.
9
J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003 3847

A R T I C L E S
Chatgilialoglu et al.
(Amberlite IRA-400) to eliminate the iron salts. The reaction mixture
was then lyophilized, and the residue was taken up in water and purified
on reverse-phase column chromatography equipped with a peristaltic
pump, a UV detector (RP18, eluted in water with a 0-30% methanol
nonlinear gradient over 5 h at a flow rate of 4 mL/min, detector at 260
nm). Two fractions were collected and lyophilized to give (5′R)-5′,8-
cyclo-dAdo (21) in 61% and a mixture of some minor products in ca.
30% overall yield. Yields are based on the recovered starting bromide.
5′,8-Cyclo-2′,5′-dideoxyAdo (23) and (5′S)-5′,8-cyclo-dAdo were
obtained as pure materials from the mixture of minor products
repeatedly chromatographed on reverse-phase silica gel and spectro-
scopically characterized.
lowest reaction barrier. The rate constants for the reactions of
the C5′ radical with O2, Fe3+, and MV2+ are similar to those
reported for the 1,2-dihydroxyethyl radical and typical of other
sugar radicals in the 2′-deoxyribo series.47 On the other hand,
the aminyl radical 3 adds reversibly to molecular oxygen and
accepts electrons only from the strongest oxidant (Fe3+). Taking
advantage of the observed reactivities, we developed a syntheti-
cally useful one-pot procedure that allows for the conversion
of 8-Br-dAdo to 5′,8-cyclo-dAdo in a diastereoisomeric ratio
(5′R):(5′S) ) 6:1.
From a biological perspective, the main reactions of the 2′-
deoxyadenin-5′-yl radical in DNA are cyclization, repair reaction
by hydrogen abstraction from glutathione, and trapping by O2
to give the corresponding peroxyl radical.6,54 Assuming the rate
constant for the cyclization of 2′-deoxyadenin-5′-yl radical in
DNA is close to 105 s-1 as seen from the present study, we
deduced that both the repair reaction by glutathione (at the
millimolar level in biological systems)55 and the trapping by
O2 (at the micromolar level) would compete with cyclization.56
However, it is worth recalling that the diastereoisomeric ratio
(5′R):(5′S) of 5′,8-cyclo-dAdo moieties in both single- and
double-stranded DNA are approximately 2.1,2 Using the data
obtained from the present study, we suggested that restricted
conformations due to the supramolecular organization should
contribute to decreasing considerably the cyclization rate
constant, due to both enthalpic and entropic effects. Indeed, it
has already been observed that the 5′,8-cyclo-dAdo moieties
decrease substantially upon irradiation of DNA in the presence
of molecular oxygen.1
Alternatively, 3 mL solutions were irradiated at different doses. The
crude reaction mixture was passed through ion-exchange resin (Am-
berlite IRA-400) to eliminate the iron salts and monitored by HPLC
on a C18-reverse-phase column (Waters XTERRA, 150 × 4.6 mm, 5
µm), eluted in water with a 0-15 acetonitrile linear gradient over 30
min at a flow rate of 0.4 mL/min and detected at 254 nm. All products
were identified and quantified by comparison with authentic samples.
For example, the experiment with 1.5 kGy dose gave the following
results in order of elution (retention time/min, yield): adenine (22.01,
5%), (5′R)-5′,8-cyclo-dAdo (22.96, 60%), (5′S)-5′,8-cyclo-dAdo (28.94,
10%), dAdo (29.92, 7%), and 5′,8-cyclo-2′,5′-dideoxyAdo (30.85, 7%).
Yields are based on the 38% consumption of 8-Br-dAdo (retention time
40.64 min).
(5′R)-5′,8-Cyclo-2′-deoxyadenosine (21). 1H NMR (400 MHz, D2O
reference peak 4.80 ppm): δ 8.18 (s, 1H, H2), 6.57 (d, 1H, J1′2′′ ) 5.2
Hz, H1′), 4.90 (d, 1H, J5′4′ ) 1.2 Hz, H5′), 4.75 (d, 1H, J4′5′ ) 1.2 Hz,
H4′), 4.46 (dd, 1H, J3′2′ ) 7.6 Hz, J3′2′′ ) 3.6 Hz, H3′), 2.59 (dd, 1H,
J2′2′′ ) -13.6 Hz, J2′3′ ) 7.6 Hz, H2′), 2.31 (dt, 1H, J2′′2′ ) -13.6 Hz,
1
J ) 4.4 Hz, H2′′). H NMR (400 MHz, DMSO-d6): δ 8.08 (s, 1H,
H2), 7.24 (s, 2H, NH2), 6.41 (d, 1H, J1′2′′ ) 4.4 Hz, H1′), 6.18 (d, 1H,
J ) 6.0 Hz, OH5′), 5.43 (d, 1H, J ) 3.6 Hz, OH3′), 4.59 (d, 1H, J )
6.0 Hz, singlet after D2O quenching, H5′), 4.43 (s, 1H, H4′), 4.19 (m,
1H, H3′), 2.26 (dd, 1H, J2′2′′ ) -13.6 Hz, J2′3′ ) 7.6 Hz, H2′), 1.99
(dt, 1H, J2′′2′ ) -12.8 Hz, J2′′1′ ) J2′′3′ ) 4.4 Hz, H2′′). ES-MS (positive
mode): m/z 250 (MH+). MS2 (250): 232, 214, 164. MS3 (164): 136.
5′,8-Cyclo-2′,5′-dideoxyadenosine (23). 1H NMR (400 MHz,
D2O): δ 8.00 (s, 1H, H-2), 6.40 (d, 1H, J1′2′′ ) 5.2 Hz, H1′), 4.85 (d,
1H, J4′5′ ) 6.4 Hz, H4′), 4.49 (dd, 1H, J3′2′ ) 7.2 Hz, J3′2′′ ) 3.2 Hz,
H3′), 3.45 (dd, 1H, J5′5′′ ) -18.4 Hz, J5′4′ ) 6.4 Hz, H5′), 3.20 (d, 1H,
J5′′5′ ) -18.4 Hz, H5′′), 2.63 (dd, 1H, J2′2′′ ) -14 Hz, J2′3′ ) 7.2 Hz,
H2′), 2.26 (ddd, 1H, J2′′2′ ) -14 Hz, J2′′1′ ) 5.2 Hz, J2′′3′ ) 3.2 Hz,
H2′′). ES-MS (positive mode): m/z 234 (MH+).
(5′S)-5′,8-Cyclo-2′-deoxyadenosine. 1H NMR (400 MHz, DMSO-
d6): δ 8.06 (s, 1H, H2), 7.18 (s, 2H, NH2), 6.35 (d, 2H, J ) 4.0 Hz,
H1′and OH5′, after D2O quenching: d, 1H, J1′2′′ ) 4.4 Hz, H1′), 5.40
(d, 1H, J ) 4.8 Hz, OH3′), 5.04 (dd, 1H, J ) 6.0, 6.4 Hz, after D2O
quenching: d, 1H, J5′4′ ) 6.0 Hz, H5′), 4.62 (m, 1H, H3′), 4.46 (d,
1H, J4′5′ ) 6.0 Hz, H4′), 2.34 (dd, 1H, J2′2′′ ) -13.2 Hz, J2′3′ ) 7.6
Hz, H2′), 2.03 (dt, 1H, J2′′2′ ) -13.2 Hz, J2′′1′ ) J2′′3′ ) 4.4 Hz, H2′′).
ES-MS (positive mode): m/z 250 (MH+).
Experimental Section
Pulse Radiolysis. Pulse radiolysis with optical absorption detection
was performed by using the 12 MeV linear accelerator, which delivered
20-200 ns electron pulses with doses between 5 and 50 Gy, by which
-
HO•, H•, and eaq are generated with 1-20 µM concentrations. The
pulse irradiations were performed at room temperature (22 ( 2 °C) on
samples contained in Spectrosil quartz cells of 2 cm optical path length.
Solutions were protected from the analyzing light by means of a shutter
and appropriate cutoff filters. The bandwith used throughout the pulse
radiolysis experiments was 5 nm. The radiation dose per pulse was
monitored by means of a charge collector placed behind the irradiation
cell and calibrated with a N2O-saturated solution containing 0.1 M
HCO2- and 0.5 mM methyl viologen, using Gꢀ ) 9.66 × 10-4 m2 J-1
at 602 nm.39 G(X) represents the number of moles of species X formed
or consumed per joule of energy absorbed by the system.
Continuous Radiolysis. Continuous radiolyses were performed at
room temperature (22 ( 2 °C) on 10-100 mL samples using a 60Co-
Gammacell, with dose rates between 20 and 25 Gy min-1. The absorbed
radiation dose was determined with the Fricke chemical dosimeter, by
taking G(Fe3+) ) 1.61 µmol J-1 57 The reactions of 8-Br-dAdo58 with
.
-
eaq and H• were investigated using deareated aqueous solutions
containing 1.5 mM substrate and 0.25 M t-BuOH in the presence or
absence of 4 mM K4Fe(CN)6 at pH ≈ 7.
Acknowledgment. We are grateful to Roman Flyunt and Rita
Bazzanini for their valuable contribution in the initial stage of
the project.23 We thank Clara Caminal i Comadira and Maria
Duca for the help in the product isolation, and Carla Ferreri
and M. Luisa Navacchia for valuable discussions. We also thank
Angelo Monti and Alessandro Martelli for assistance with pulse
radiolysis experiments. We are also grateful to Francesco
Tortorella of Bruker Daltonics S.r.l. for the Ion Trap MSn runs
on an Esquire 3000 plus instrument.
One liter of solution was irradiated with a total dose up to 3 kGy.
The crude reaction mixture was passed through ion-exchange resin
(55) (a) The rate constant for the reaction of CH3CH(•)OH with glutathione is
1.1 × 108 M-1 s-1 25,30
.
(b) The intracellular level of glutathione in
mammalian cells is in the 0.5-10 mM range, see: Meister, A.; Anderson,
M. E. Annu. ReV. Biochem. 1983, 52, 711-760.
(56) The oxygen concentration is low in the nucleus, see: Zander, R. Z.
Naturforsch. 1976, 31C, 339-352. Zander, R. AdV. Exp. Med. Biol. 1976,
75, 463-467.
(57) Spinks, J. W. T.; Woods, R. J. An Introduction to Radiation Chemistry,
3rd ed.; Wiley: New York, 1990; p 100.
(58) 8-Br-dAdo was purchased from Berry&Associates, Inc.
JA029374D
9
3848 J. AM. CHEM. SOC. VOL. 125, NO. 13, 2003