Interstrand cross-link formation in duplex and triplex DNA by modified pyrimidines

Article abstract of DOI:10.1021/ja802177u

DNA interstrand cross-links have important biological consequences and are useful biotechnology tools. Phenylselenyl substituted derivatives of thymidine (1) and 5-methyl-2′-deoxycytidine (5) produce interstrand cross-links in duplex DNA when oxidized by NaIO₄. The mechanism involves a [2,3]-sigmatropic rearrangement of the respective selenoxides to the corresponding methide type intermediates, which ultimately produce the interstrand cross-links. Determination of the rate constants for the selenoxide rearrangements indicates that the rate-determining step for cross-linking is after methide formation. Cross-linking by the thymidine derivative in duplex DNA shows a modest kinetic preference when flanked by pyrimidines as opposed to purines. In contrast, the rate constant for cross-link formation from 5 opposite dG in duplex DNA is strongly dependent upon the flanking sequence and, in general, is at least an order of magnitude slower than that for 1 in an otherwise identical sequence. Introduction of mispairs at the base pairs flanking 5 or substitution of the opposing dG by dI significantly increases the rate constant and yield for cross-linking, indicating that stronger hydrogen bonding between the methide derived from it and dG compared to dA and the respective electrophile derived from 1 limits reaction by increasing the barrier to rotation into the required syn-conformation. Incorporation of 1 or 5 in triplex forming oligonucleotides (TFOs) that utilize Hoogsteen base pairing also yields interstrand cross-links. The dC derivative produces ICLs approximately 10x faster than the thymidine derivative when incorporated at the 5′-termini of the TFOs and higher yields when incorporated at internal sites. The slower, less efficient ICL formation emanating from 1 is attributed to reaction at N1-dA, which requires local melting of the duplex. In contrast, 5 produces cross-links by reacting with N7-dG. The cross-linking reactions of 1 and 5 illustrate the versatility and utility of these molecules as mechanistic probes and tools for biotechnology.

Full text of DOI:10.1021/ja802177u

Published on Web 07/12/2008
Interstrand Cross-Link Formation in Duplex and Triplex DNA
by Modified Pyrimidines
Xiaohua Peng,† In Seok Hong,† Hong Li,‡ Michael M. Seidman,‡ and
Marc M. Greenberg*,†
Department of Chemistry, Johns Hopkins UniVersity, 3400 North Charles Street, Baltimore,
Maryland 21218 and Laboratory of Molecular Gerontology, National Institute on Aging,
National Institutes of Health, 5600 Nathan Shock DriVe, Baltimore, Maryland 21224
Received March 24, 2008; E-mail: mgreenberg@jhu.edu
Abstract: DNA interstrand cross-links have important biological consequences and are useful biotechnology
tools. Phenylselenyl substituted derivatives of thymidine (1) and 5-methyl-2′-deoxycytidine (5) produce
interstrand cross-links in duplex DNA when oxidized by NaIO4. The mechanism involves a [2,3]-sigmatropic
rearrangement of the respective selenoxides to the corresponding methide type intermediates, which
ultimately produce the interstrand cross-links. Determination of the rate constants for the selenoxide
rearrangements indicates that the rate-determining step for cross-linking is after methide formation. Cross-
linking by the thymidine derivative in duplex DNA shows a modest kinetic preference when flanked by
pyrimidines as opposed to purines. In contrast, the rate constant for cross-link formation from 5 opposite
dG in duplex DNA is strongly dependent upon the flanking sequence and, in general, is at least an order
of magnitude slower than that for 1 in an otherwise identical sequence. Introduction of mispairs at the base
pairs flanking 5 or substitution of the opposing dG by dI significantly increases the rate constant and yield
for cross-linking, indicating that stronger hydrogen bonding between the methide derived from it and dG
compared to dA and the respective electrophile derived from 1 limits reaction by increasing the barrier to
rotation into the required syn-conformation. Incorporation of 1 or 5 in triplex forming oligonucleotides (TFOs)
that utilize Hoogsteen base pairing also yields interstrand cross-links. The dC derivative produces ICLs
approximately 10× faster than the thymidine derivative when incorporated at the 5′-termini of the TFOs
and higher yields when incorporated at internal sites. The slower, less efficient ICL formation emanating
from 1 is attributed to reaction at N1-dA, which requires local melting of the duplex. In contrast, 5 produces
cross-links by reacting with N7-dG. The cross-linking reactions of 1 and 5 illustrate the versatility and utility
of these molecules as mechanistic probes and tools for biotechnology.
DNA alkylation is a common chemical process that is useful
as a chemical tool but can also have biological consequences.
For instance, alkylating agents are employed in the laboratory
for determining DNA structure. However, alkylated DNA can
be genotoxic and/or cytotoxic.1 In addition, a number of
molecules, including some antitumor agents, react with DNA
by forming covalent bonds to the opposing strands.2 The cross-
linked strands are strong blocks to replication and are cytotoxic
if unrepaired.3 Modified oligonucleotides containing alkylating
agents have also been used to form interstrand cross-links in
duplex and triplex DNA.4–6 We recently described a modified
thymidine (1) that forms stable base pairs with dA and produces
interstrand cross-links (ICLs) (4a, 4b) in duplex DNA under
mild oxidative conditions (Scheme 1).7–9 The cross-linking
reaction is unusual in that like the antitumor agent N,N′-bis(2-
chloroethyl)-nitrosourea (BCNU) the cross-links are produced
between opposing nucleotides in the duplex.10 The phenyl
selenide (1) has been used to sensitize DNA to the effects of
γ-radiolysis and as a tool for detecting single nucleotide
polymorphisms.11,12 This efficient cross-linking chemistry has
now been extended to produce cross-links between dG and a
derivative of 2′-deoxycytidine (5). Mechanistic characterization
of the reactivity of 1 and 5 in duplex and triplex DNA is
described.
† Johns Hopkins University.
‡ National Institute on Aging.
(1) Gates, K. S.; Nooner, T.; Dutta, S. Chem. Res. Toxicol. 2004, 17,
839–856.
(7) Hong, I. S.; Greenberg, M. M. J. Am. Chem. Soc. 2005, 127, 3692–
3693.
(2) Veldhuyzen, W. F.; Parde, P.; Rokita, S. E. J. Am. Chem. Soc. 2003,
125, 14005–14013.
(8) Hong, I. S.; Greenberg, M. M. J. Am. Chem. Soc. 2005, 127, 10510–
10511.
(3) Noll, D. M.; Mason, T. M.; Miller, P. S. Chem. ReV. 2006, 106, 277–
301.
(9) Hong, I. S.; Ding, H.; Greenberg, M. M. J. Am. Chem. Soc. 2006,
128, 485–491.
(4) Zhou, Q.; Rokita, S. E. Proc. Nat. Acad. Sci. U.S.A. 2003, 100, 15452–
15457.
(10) Fischhaber, P. L.; Gall, A. S.; Duncan, J. A.; Hopkins, P. B. Cancer
Res. 1999, 59, 4363–4368.
(5) Nagatsugi, F.; Kawasaki, T.; Usui, D.; Maeda, M.; Sasaki, S. J. Am.
Chem. Soc. 1999, 121, 6753–6754.
(11) Hong, I. S.; Ding, H.; Greenberg, M. M. J. Am. Chem. Soc. 2006,
128, 2230–2231.
(6) Kawasaki, T.; Nagatsugi, F.; Ali, M. M.; Maeda, M.; Sugiyama, K.;
Hori, K.; Sasaki, S. J. Org. Chem. 2005, 70, 14–23.
(12) Peng, X.; Greenberg, M. M. Nucleic Acids Res. 2008, 36, e31.
9
10.1021/ja802177u CCC: $40.75
2008 American Chemical Society
J. AM. CHEM. SOC. 2008, 130, 10299–10306 10299

A R T I C L E S
Peng et al.
Scheme 1. Interstrand Cross-Link Formation from 1 under Oxidative Conditions
Scheme 2. Synthesis of Phosphoramidite 10a
1
as H2O2 and O2. Cross-links to dA are produced in 40-50%
yield with a half-life of ∼30 min at room temperature.
Examination of molecular models suggested that the analogous
phenyl selenide derivative of 5-methyl-2′-deoxycytidine (5)
would cross-link an opposing dG when oxidized to the phenyl
selenoxide. Similar qualitative examination of 1:dA and 5:dG
Hoogsteen base pairs suggested that the modified nucleotides
would also be useful for producing interstrand cross-links (ICLs)
in triple helical DNA. The investigations described herein reveal
that these predictions were borne out, but the chemistry is more
complex than models suggest.
a (a) POCl3, triazole, TEA, CH3CN; (b) Concentrated aqueous NH4OH,
dioxane; (c) Ac2O, DMF; (d) DMTCl, pyridine; (e) N,N′-diisopropyl
2-cyanoethyl phosphoramidic chloride, N,N′-diisopropylethylamine, CH2Cl2.
In general, small molecules and modified oligonucleotides
that cross-link DNA are useful tools for mechanistic studies
and in biotechnology applications.13–17 Oligonucleotide cross-
linking is also potentially useful in therapeutic applications.18–21
A desirable oligonucleotide cross-linking reaction should pro-
ceed in high yield, with good selectivity and a high rate constant.
The phenyl selenide nucleotide (1) satisfies these requirements.
It undergoes transformation into the selenoxide under a variety
of conditions, including with oxidants as mild as NaIO4. Cross-
linking with the opposing dA in duplex DNA is also induced
using reactive oxygen species that are produced in cells, such
Results and Discussion
Synthesis of the Cross-Linking Precursor and Its
Incorporation into Oligonucleotides. The synthesis of oligo-
nucleotides containing 1 was reported.7 The monomeric precur-
sor to the methide (8) and respective phosphoramidite (10) for
preparing oligonucleotides containing 5 was synthesized in
routine fashion from 6 (Scheme 2).9 Following synthesis of the
triazole (7), the exocyclic amine was introduced and the acetate
protecting groups were cleaved in one pot. A variety of
approaches were taken to carry 8 forward. Ultimately, the amine
(8) was protected as its acetate without protecting the hydroxyl
groups, as it was not advantageous to transiently protect the
hydroxyl groups. Good yields of 9 were obtained by carrying
out tritylation on the crude acetylated nucleoside. The phos-
phoramidite was prepared from 9 using standard conditions.
Oligonucleotides containing 5 were prepared using com-
mercially available ꢀ-cyanoethyl phosphoramidites containing
phenoxyacetyl protecting groups on the exocyclic amines of dA
and dG. The exocyclic amine of dC was protected with acetate.
Due to the potential oxidation of 5, t-BuOOH was used on the
synthesizer in place of I2/pyridine/H2O. In addition, the coupling
time for 5 was extended to 5 min. The oligonucleotides
containing 5 were deprotected using concentrated aqueous
ammonium hydroxide at 25 °C for 3 h. Denaturing polyacry-
lamide gel electrophoresis was used to separate the oligonucle-
otides from deletion products. The oligonucleotides were further
purified by reversed phase HPLC to separate the respective
(13) Povsic, T. J.; Strobel, S. A.; Dervan, P. B. J. Am. Chem. Soc. 1992,
114, 5934–41.
(14) Ferentz, A. E.; Keating, T. A.; Verdine, G. L. J. Am. Chem. Soc. 1993,
115, 9006–14.
(15) Osborne, S. E.; Cain, R. J.; Glick, G. D. J. Am. Chem. Soc. 1997,
119, 1171–1182.
(16) Coleman, R. S.; Pires, R. M. Nucleic Acids Res. 1997, 25, 4771–
4777.
(17) Glick, G. D. In Current Protocols in Nucleic Acid Chemistry;
Beaucage, S. L. Ed.; John Wiley & Sons, Inc.: New York, 2003; Vol.
2, pp 5.7.1-5.7.13.
(18) Li, H.; Broughton-Head, V. J.; Peng, G.; Powers, V. E. C.; Ovens,
M. J.; Fox, K. R.; Brown, T. Bioconjugate Chem. 2006, 17, 1561–
1567.
(19) Puri, N.; Majumdar, A.; Cuenoud, B.; Natt, F.; Martin, P.; Boyd, A.;
Miller, P. S.; Seidman, M. M. Biochemistry 2002, 41, 7716–7726.
(20) Puri, N.; Majumdar, A.; Cuenoud, B.; Miller, P. S.; Seidman, M. M.
Biochemistry 2004, 43, 1343–1351.
(21) Nagatsugi, F.; Sasaki, S.; Miller, P. S.; Seidman, M. M. Nucleic Acids
Res. 2003, 31, e31.
9
10300 J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008

DNA Interstrand Cross-Link Formation
A R T I C L E S
Table 1. Comparison of UV-Melting Temperatures of Duplexes
Containing 5 or MedC
Scheme 3. Oxidative Rearrangement of 8
5′-d(AGA TGG AT X TAG GAC G)
3′-d(TCT ACC TA9N A7Tc CTG C)
duplex
X
N
TM (°C)
11a
11b
11c
11d
5
5
G
A
G
A
58.0 ( 0.6
49.2 ( 0.3
63.3 ( 0.7
49.6 ( 0.8
MedC
MedC
5, respectively. The resonances corresponding to 8 were replaced
by those that were indicative of a diastereomeric mixture of 13
and a second molecule identified as the selenoxide (12), which
was also formed as a mixture of diastereomers (Scheme 3). The
vinyl protons (∼δ 5.62, 5.86) and the C6-proton (∼ δ 6.03)
were used to identify methide 13. Evidence for the selenoxide
(12) was gleaned from the appearance of two C1′-protons (∼δ
6.14, 6.04) and two C6-vinyl protons (∼δ 7.30, 7.21). After
incubating the sample at 37 °C for 4 h (Figure 1B), the
resonances corresponding to the methide (13) and 12 were
replaced by those of the C5-hydroxymethyl product (HOMedC)
(Figure 1C). Comparing these data to those obtained upon
subjection of the thymidine substrate (1) to the same reaction
conditions suggests that the 2′-deoxycytidine selenoxide (12)
undergoes the [2,3]-sigmatropic rearrangement more slowly than
the respective thymidine analogue (2) for which there was no
1H NMR evidence for the respective selenoxide 10 min after
adding the oxidant.8
Absorption spectroscopy was used to quantitatively compare
the [2,3]-sigmatropic rearrangements of the thymidine (2) and
2′-deoxycytidine selenoxides (12). The disappearance of the
selenoxides was followed continuously by taking advantage of
the blue-shifted absorption spectra of the rearrangement products
(3, 13). The selenoxides followed first-order decay.22 However,
2 ((6.1 ( 0.8) × 10-3 s-1, t1/2 ) 1.9 min) rearranged
approximately 5 times more rapidly than the 2′-deoxycytidine
(12) compound ((1.2 ( 0.1) × 10-3 s-1, t1/2 ) 9.6 min).
Interstrand Cross-Linking by 5. Reaction of 11a or 14 with
NaIO4 (5 mM) at 37 °C produced cross-linked product in as
high as 47.0 ( 0.5% yield, and the rate of ICL formation at 37
Figure 1. 1H NMR analysis of oxidative rearrangement of 8 in deuterated
phosphate buffer: (A) 10 min after addition of NaIO4; (B) 4 h after addition
of NaIO4. (C) Independently synthesized HOMedC.
phenyl selenide containing oligonucleotides from the hydroxy-
methyl (e.g., HOMedC) degradation product.
°C followed first-order kinetics (kICL ) (1.1 ( 0.2) × 10-4 s-1
;
t1/2 ) 104 min). Cross-linking from 5 is approximately 1 order
of magnitude slower than that from the thymidine analogue (1)
when it too is flanked by thymidines (kICL ) (1.2 ( 0.3) ×
10-3 s-1; t1/2 ) 9.4 min).12 The molecular weight of the cross-
linked product from 11a was determined using ESI-MS.22 This
revealed that O2 was not incorporated and that the product
resulted from the net loss of benzene selenol from the starting
duplex. Insight into the site of electrophilic attack on the
opposite strand and the structure of the product was obtained
by subjecting duplexes containing 2′-deoxyinosine (dI, 15 and
16) or dA (11b) opposite 5 to the reaction conditions. Although
cross-links formed in modest yield when dA (8.2 ( 0.3%) was
mispaired opposite 5, the yield of cross-link in a duplex (16)
containing dI (52.1 ( 0.5%) was comparable to when dG was
present. ESI-MS revealed that the cross-linking product when
dI opposed the phenyl selenide (16) also arose from the formal
elimination of benzene selenol.22
After characterizing the oligonucleotides containing 5 by ESI-
MS, the effect of the phenyl selenide on duplex thermal stability
was analyzed by measuring the UV-melting temperatures (TM,
Table 1). The phenyl selenide reduced the TM 5.3 °C relative
to when 5-methyl-2′-deoxycytidine (MedC) was present, which
is comparable to the depression observed when 1 was substituted
for thymidine opposite dA.7 The effect of a mispair opposite 5
was greater than when 1 was examined, but the reduction was
smaller than when dA was placed opposite MedC. Overall, these
data suggest that 5 forms a Watson-Crick base pair with dG,
albeit one that is less stable than a MedC:dG pair.
Oxidative Transformation of the Phenyl Selenides into
Methide Species. Reaction of 8 with NaIO4 in deuterated
phosphate buffer was analyzed using 1H NMR (Figure 1).
Analysis 10 min after addition of periodate revealed that starting
material (8) was completely consumed (Figure 1A), as evidenced
by the absence of diagnostic resonances at δ 6.87 (singlet) and
δ 5.88 (triplet) corresponding to the vinyl and C1′-protons of
(22) See Supporting Information.
9
J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008 10301

A R T I C L E S
Peng et al.
significant effect on cross-linking even though NaIO4 induced
cross-linking reactions were run at 37 °C instead of 25 °C.
Although 5 effectively yielded cross-links when flanked by dT
(per above), only 4.0 ( 0.3% ICL was observed when the
phenyl selenide was flanked by dA (19). Moreover, no cross-
linked product was observed when 5 was surrounded by dC
(20) or dG (21) at 37 °C. Because 13 must adopt the
syn-conformation in order to react with the opposing strand,
local duplex stability was suspected to be the cause of the
sequence effect on cross-linking. We postulated that the more
stable base pairing with dG than the thymidine analogue
opposite dA, as well as flanking sequence effects on duplex
stability, affected cross-linking by 5 (and 13). The latter is
consistent with the inverse correlation between calculated TM’s
(dT < dA < dC < dG)25 for the respective duplexes containing
dC instead of 5 and the observed sequence flanking dependence
(dT > dA > dC ≈ dG) on cross-linking.
The nucleotide position(s) at which cross-linking occurred
was determined using Hopkins’ hydroxyl radical cleavage
method (Figure 2).22,23 As expected, all cross-links in 11a
emanated from the position at which 5 is produced (Figure 2A).
Analysis of the position of the cross-link in the complementary
strand revealed that although the majority of the cross-links
involve reaction with the dG opposite 5, 15-20% of the cross-
links involve dA9 (Figure 2B). In contrast, labeling the
complementary strand of 11a at its 3′-terminus clearly showed
that cross-linking did not occur at dA7 (Figure 2C). The lower
cross-linking fidelity exhibited by 11a than the thymidine
analogue (1)7 was attributed to the reduced nucleophilicity of
the opposing nucleotide (dG) and not the lower reactivity of
the methide (13) produced from 5. Kinetic experiments and
additional product characterization described below attest to this.
The similarity in cross-linking yields between duplexes
containing dG or dI opposite the methide (13) suggested that
the electrophile reacted with the N1-position of the purines. This
hypothesis was verified upon characterization of the product
formed from reaction with dI. Initially, 17 was isolated from a
reaction between monomeric 5 and dI. 1H NMR characterization
of 17 was consistent with the proposed structure, as evidenced
for example by the vinyl proton pattern.22 Formation of 17 in
DNA was confirmed by HPLC coinjection, following enzymatic
digestion of 16 subjected to NaIO4 treatment. The analogous
product (18) from reaction between 13 and dG could not be
isolated in pure form in sufficient quantities for NMR structural
analysis. However, ESI-MS was consistent with its formation.
In contrast to reaction of the thymidine methide (3) with dA,
the cross-linked products formed in duplexes in which 13 was
opposed by dG or dI were stable upon incubation in the presence
of NaN3 (0.3 M) at 37 °C for 5 days. This also is consistent
with alkylation at the N1 position of dI or dG by 13, which
should give rise to a stable product.24 Reaction at the N2-amino
of dG would also give rise to a stable product. Although we
cannot rule out reaction at this position, the similarities between
reaction with dI and dG suggest that reaction at the N2-amino
does not occur.
Additional experimental support for the hypothesis that
increased base pair stability in the vicinity of 5 adversely
affected its ability to react with a nucleotide in the opposing
strand was obtained by investigating temperature and sequence
effects on cross-linking. Increasing the reaction temperature from
37 to 45 °C increased cross-linking in 19 from 4.0 ( 0.3% to
10.4 ( 0.3% and from <1% to 8.7 ( 0.6% in 20. Moreover,
introducing mismatches at the flanking positions resulted in even
more significant increases in cross-linking yields at 37 °C (Table
2). The most profound evidence in support of the effect of
duplex thermal stability on cross-linking from DNA containing
5 was obtained by substituting dI (15, 25), which forms only
two hydrogen bonds with 5 (and 13), for dG opposite the phenyl
selenide. High cross-link yields (29.0 ( 1.2%) were observed
at 37 °C when dI was flanked by dG (25), whereas a comparable
sequence yielded <1% ICL product when 5 was opposite dG
(20). In addition, the rate constant for cross-link formation in
25 (kICL ) (1.9 ( 0.3) × 10-4 s-1; t1/2 ) 61.8 min; 37 °C) was
even greater than when 5 was opposite dG and flanked by
thymidines (14). However, the rate constant for cross-linking
in 25 was ∼10-times slower at the same temperature than when
the thymidine analogue (1) was flanked by dC (kICL ) (1.5 (
0.2) × 10-3 s-1; 37 °C).12
γ-Radiolysis Mediated Cross-Linking. We previously showed
that the major pathway for phenyl selenide 1 cross-linking upon
γ-radiolysis proceeded through the selenoxide (Scheme 1).11
Consequently, cross-linking due to 137Cs irradiation of duplex
DNA containing 5 should parallel the oxidative reaction induced
by NaIO4. Indeed, the same local sequence effects on cross-
linking by γ-radiolysis were detected as when the substrates
were treated with NaIO4. For instance, no cross-linked product
was observed when duplexes containing 5 that are flanked by
dG or dC were irradiated (data not shown). Furthermore, the
(23) Millard, J. T.; Weidner, M. F.; Kirchner, J. J.; Ribeiro, S.; Hopkins,
P. B. Nucleic Acids Res. 1991, 19, 1885–1892.
(24) Weinert, E. E.; Frankenfield, K. N.; Rokita, S. E. Chem. Res. Toxicol.
2005, 18, 1364–1370.
Flanking Sequence Effects on Cross-Linking in Duplex
DNA Containing 5. In contrast to the reactivity of the thymidine
analogue (1), we found that the sequence flanking 5 had a
(25) HyTher is available at: http://ozone2.chem.wayne.edu/Hyther/
hytherm1main.html
9
10302 J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008

DNA Interstrand Cross-Link Formation
A R T I C L E S
Table 2. Effect of Flanking Sequence Mismatches on ICL
Formation from 5
5′-d(TCG TAC CAA 5AA CCA GTC G)
3′-d(AGC ATG GTN GNT GGT CAG C)
duplex
N
% ICL
19
22
23
24
T
G
C
A
4.0 ( 0.3
18.3 ( 0.4
38.2 ( 2.6
49.6 ( 5.6
stabilized at slightly acidic pH. Stable triplexes can be formed
at physiological pH using synthetically modified nucleotides in
the TFO.19–21,26,27 Since our goal was to establish the viability
of 1 and 5 as triplex cross-linking agents, for synthetic ease we
opted to construct the TFOs using thymidine and 2′-deoxycy-
tidine. This required that we determine that the rearrangement/
alkylation chemistry function at pH 5.5 and that triplexes
containing the modified pyrimidines be sufficiently stable under
these conditions as well.
The required cross-linking reaction temperature was deter-
mined by measuring the TM’s of triplexes containing 1 and 5 at
the 5′-termini or an internal position of a TFO at pH 5.5 (Table
3). In each instance, the triplex containing the modified
pyrimidine at an internal position was less stable than that at
the 5′-terminus. The lowest TM was 34 °C, leading us to carry
out the cross-linking reactions at 25 °C. The compatibility of
the ICL process at pH 5.5 (Hepes buffer) was established by
examining cross-link formation in duplex DNA. Reducing the
pH from 7.2 to 5.5 decreased the ICL yield (25 °C) from 1 in
30 to 19 ( 0.4% from 30.0 ( 1.5%.
These conditions were utilized for determining the NaIO4 (5
mM) mediated cross-linking of 26-29 at 25 °C (Table 4). No
increase in ICL yields was observed when the Mg2+ concentra-
tion was increased to 50 mM from 10 mM. ESI-MS analysis
of isolated product from reactions of 31 and 29 confirmed the
expected coupling between the TFO and the polypurine strand.22
Comparable cross-link yields were obtained when either modi-
fied pyrimidine was incorporated at the 5′-termini of the
respective TFOs. In addition, the ICL yields in the triplexes
were comparable to those measured in the model duplex (30)
under the same pH conditions. In contrast, the ICL yield from
1 decreased significantly when it was incorporated at an internal
position (26), whereas the reactivity of 13 was unaffected by
its position in the oligonucleotide. Although the ICL yields
obtained from the TFOs containing 1 or 5 at their 5′-termini
were comparable, the rate constant for cross-link formation from
5 in 29 (kICL ) (8.6 ( 0.1) × 10-4 s-1; 25 °C) was ∼10-fold
higher than that from 1 in a comparable triplex (27, kICL ) (8.2
( 0.8) × 10-5 s-1; 25 °C). The relative reactivity of 1 (3) and
5 (13) in triplex DNA is in marked contrast to what was
observed in comparable duplexes. The rate constant for cross-
Figure 2. Determination of interstrand cross-linking position via OH•
cleavage of NaIO4 treated 11a: (A) 5′-32P-11a, phenyl selenide containing
strand labeled; (B) 5′-32P-11a, complementary strand labeled; (C) 3′-32P-
11a, complementary strand labeled.
rate constants for cross-link formation in 14 (kICL ) (1.3 ( 0.1)
× 10-4 s-1; t1/2 ) 90.9 min; 37 °C) and 25 (kICL ) (1.7 ( 0.2)
× 10-4 s-1; t1/2 ) 69 min; 37 °C) upon exposure to γ-radiolysis
were within experimental error of those measured using NaIO4.
These experiments not only further verify that cross-linking
involves selenoxide formation (12) but also reveal that 5 (unlike
1) will not be useful as a radiosensitizing agent.11
Interstrand Cross-Linking in Triplexes. Inspection of mo-
lecular models suggested that incorporating 1 or 5 in a triplex
forming oligonucleotide (TFO) that utilizes Hoogsteen base
pairing could produce ICLs via the same oxidation/rearrange-
ment mechanism (Scheme 4). The proximity of the electrophile
in syn-3 or syn-13 to the nucleophilic N7-position of the purines
suggested this position as the likely site of attack when the
modified nucleotides adopt the syn-conformation. Hoogsteen
triplexes are less stable than Watson-Crick duplexes and are
(26) Singleton, S. F.; Dervan, P. B. Biochemistry 1992, 31, 10995–1003.
(27) Povsic, T. J.; Dervan, P. B. J. Am. Chem. Soc. 1989, 111, 3059–61.
9
J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008 10303

A R T I C L E S
Peng et al.
Scheme 4. Alkylation in Triplex DNA via Hoogsteen Base Pair Formation
Table 3. UV-Melting Temperatures of Triplexes
Although these properties help rationalize why 5 is more
effective overall at forming ICLs in the triplexes, it does not
explain the effect of position in the TFO on the reactivity of
the TFOs containing 1.
In order to provide further insight into the reactivity of the
TFOs containing 1 and 5, we examined the stability and
reactivity of the ICL products. The stability and alkaline lability
of DNA alkylation products depend upon the reaction site. The
cross-links formed from 1 were almost completely destroyed
upon heating at 90 °C (pH 7.2) for 25 min, which is consistent
with reaction at N7- or N1- of dA, but ruled out reaction of 3
with the N6-amino group. In addition, reaction at N3-dA was
considered unlikely, as that nucleophile is in the minor groove.
The cross-linked products derived from 5 were also almost
completely decomposed upon heating, which is consistent with
reaction occurring at the N7-position of dG as well. In contrast,
the cross-links derived from 1 and 5 exhibited different levels
of stability to 1 M piperidine treatment (90 °C, 20 min). Mild
alkaline hydrolysis of N7-alkylated purines produces formami-
dopyrimidines, which are labile to the harsher piperidine
conditions.29 Hence, if N7-dA and N7-dG were the sites attacked
by 3 and 13, respectively, the cross-links should decompose
upon treatment with piperidine. Surprisingly, piperidine treat-
ment of the ICLs derived from 1 in 27 produced ∼50%
cleavage. In addition, a small amount of a product that migrated
even faster than γ-32P-ATP in 20% denaturing PAGE was
observed.22 This product was also observed when cross-linked
29 was treated with piperidine. In addition, the ICL formed from
29 was almost completely destroyed by piperidine treatment.
The formation of a radioactive product that migrated even more
rapidly than γ-32P-ATP suggested that 3 and 13 reacted with
the 5′-terminal dG in the purine strand of the duplexes (27 and
29). N7-Alkylation of the terminal dG would yield the alkylated
formamidopyrimidine under the alkaline conditions. Concomi-
tant cleavage of this lesion would ultimately result in the release
of the 5′-terminal 32P-phosphate (Figure 3). Alternatively,
deglycosylation of the N7-alkylated product would yield an
abasic site directly, which would also release the radiolabeled
phosphate upon alkaline hydrolysis.
Table 4. Interstrand Cross-Link (ISC) Yields from 1 and 5 in Triple
Helical DNA
% yield ICLa
c
triplex (phenyl selenide)
NFTb
∆
Piperidined
26 (1)
27 (1)
28 (5)
29 (5)
31 (1)
32 (5)
4.0 ( 0.1
20.0 ( 0.5
24.0 ( 0.3
25.0 ( 0.5
13.0 ( 1.0
11.5 ( 0.5
n.d.
2.5 ( 0.6
n.d.
2.2 ( 0.4
3.0 ( 0.1
1.0 ( 0.1
3.6 ( 0.3
10.0 ( 1.0
1.0 ( 0.1
1.3 ( 0.2
12.0 ( 0.5
1.0 ( 0.1
a % Yields are the average of at least two independent experiments.
b NFT ) no further treatment after reaction with NaIO4. c 90 °C for 20
min at pH 7.2. d 1.0 M Piperidine, 90 °C, 20 min.
linking from 1 situated at the 5′-terminus of the TFO was ∼20-
fold slower in the triplex than in duplex DNA, whereas the rate
constant for ICL formation from 5 increased approximately
8-fold.
The differences in reactivity between 1 and 5 in triplex DNA
were contrary to expectations based upon the anticipated
respective barriers to rotation about their glycosidic bonds.
However, the 2′-deoxycytidine derivative should be more
electrophilic under the conditions if it is protonated. In addition,
reaction at the N7-purine position of dG should be more facile
due to its greater nucleophilicity compared that of dA.28
(28) Bloomfield, V. A.; Crothers, D. M.; Tinoco, I., Jr. Nucleic Acids
Structures, Properties, and Functions; University Science Books:
Sausalito, CA, 2000.
(29) Haraguchi, K.; Delaney, M. O.; Wiederholt, C. J.; Sambandam, A.;
Hantosi, Z.; Greenberg, M. M. J. Am. Chem. Soc. 2002, 124, 3263–
3269.
9
10304 J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008

DNA Interstrand Cross-Link Formation
A R T I C L E S
Figure 3. Phosphate release upon alkaline treatment following N7-alkylation of the 5′-terminal dG in triple helical DNA. (A) Cartoon illustration of alkylation
of the 5′-terminal dG by 13. (B) Phosphate release upon alkaline treatment (* ) 32P).
Scheme 5. Effects of Alkali Treatment and Heat on N1-dA:3 Cross-Link
The hypothesis that the methides produced at the 5′-termini
of TFOs could react at a distal dG was tested by substituting
thymidine for the 5′-terminal dG in triplexes 27 and 29. Indeed,
subjection of cross-linked 32 to either heating or piperidine
resulted in almost complete destruction of the product. However,
free phosphate was not detected following alkaline treatment.
These observations suggest that 13 produced in triplex 32 did
not react at the 5′-terminal thymidine. The lability of the ICL
to alkaline conditions or heat in phosphate buffer is consistent
with reaction at N7-of the opposing dG, as predicted by simple
modeling of the Hoogsteen pairing (Scheme 4). Incorporation
of thymidine at the 5′-terminus of the polypurine-containing
strand in 31 also clarified the reactivity of 3 in triple helical
DNA. As in the experiments with 26 and 27, heating cross-
linked 31 destroyed almost all of the ICL (Table 4). However,
reaction with piperidine had essentially no effect on the ICL
produced in 31. These data suggest that although alkylation of
the opposing dA by 3 produces a product with a labile glycosidic
bond, the resistance to cleavage by alkali indicates that N7-dA
is not the position attacked because the respective formami-
dopyrimidine formed under these conditions should be labile
to piperidine.29 Of the other nitrogens on dA, only reaction at
N1 would produce a product with the observed chemical
behavior due to its ability to isomerize to an N6-alkylamino
derivative (Scheme 5).7,30,31 Although reaction at N1-dA is
unexpected based upon inspection of molecular models, this
nitrogen is more nucleophilic than the N7-dA.28 Reaction at
N1-dA requires that the Watson-Crick duplex undergo local
melting. This conformational restriction helps explain why the
rate constant for cross-linking by 3 is so much slower in the
triplex than in the duplex and even slower than the reaction of
13. It may also help explain why the ICL yield in 26 is so much
lower than that in 27. Placing the phenyl selenide at the 5′-
terminus of the TFO enables it to react with the dG of the duplex
not involved in triplex formation. In addition, fraying of the
TFO in triplex 27 facilitates the required conformational change
of the opposing dA. In comparison, the internal methide (3) in
26 may in effect be more rigid, making it more difficult for the
dA to undergo the rotation necessary to enable N1 to react with
the electrophile.
Conclusions
Oligonucleotides containing phenyl substituted derivatives of
thymidine (1) and 5-methyl-2′-deoxycytidine (5) form stable
Watson-Crick duplexes and triplexes via Hoogsteen base pairs.
The phenyl selenides undergo rapid oxidation to selenoxides
(2, 12), followed by [2,3]-sigmatropic rearrangements to elec-
9
J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008 10305

A R T I C L E S
Peng et al.
trophilic methides (3, 13). The rearrangements are facile at 25
°C, with the slower one (12 f 13) occurring with a < 10 min
half-life. Adoption of the respective syn-conformations enables
3 and 13 to cross-link with the opposing strand. Interstrand
cross-link formation from 5 (13) in duplex DNA is significantly
affected by a flanking sequence when it is opposite dG. Varying
temperature, introducing mismatches, and most significantly
replacing the opposing dG with dI reveal that stronger hydrogen
bonding between 5 (13) and dG compared to 1 (3) and dA
increases the importance of the flanking sequence (and its
commensurate effect on TM) on cross-linking, presumably by
increasing the barrier faced by 13 to adopt the syn-conformation.
While this limits the utility of 5 as a cross-linking agent in
duplex DNA, it opens the possibility of using it in other
applications. For instance, the presence of the methide in the
majorgroovepotentiallyprovidesatoolforprobingprotein-DNA
cross-linking.
The barriers to conformational isomerization do not distin-
guish the reactivity of 1 and 5 from one another when they are
incorporated into triplex forming oligonucleotides (TFOs) that
recognize purines via Hoogsteen base pairing. The syn-isomers
of the respective methides are well positioned to react at N7 of
the opposing purine. The 2′-deoxycytidine derivative (5)
produces cross-links via reaction at this position with the
opposing dG. In contrast, the greater nucleophilicity of the N1-
position of adenine overcomes the requirement to break its
hydrogen bonds with the opposing dT to react with 3 in the
TFO. Although cross-linking from 5 occurs equally well when
it is incorporated at terminal or internal positions of the TFO,
the required local melting of the duplex to form interstrand
cross-links with 3 limits the yield of this covalent modification
when the modified nucleotide is incorporated at an internal
position of the TFO. In addition, the triple helix cross-linking
reactions were carried out under nonoptimal pH conditions for
ICL formation in order to stabilize the triplexes. Based upon
our results in duplexes, we expect that higher ICL yields will
be obtained using modified TFOs that are compatible with
carrying out the experiments at physiological pH. Overall, the
phenyl selenide derivatives provide a means for efficient cross-
linking under mild reaction conditions. In addition, the require-
ment of a mild oxidant allows for temporal control of the cross-
linking reaction.
Acknowledgment. We are grateful for support of this research
by the National Institute of General Medical Sciences (GM-054996).
This research was supported in part by the Intramural Research
Program of the NIH, National Institute on Aging.
Supporting Information Available: Complete description of
experimental procedures. NMR spectra of synthetic molecules,
ESI-MS of cross-linked products, UV-vis spectra describing
selenoxide rearrangements, HPLC analysis of enzyme digests,
and autoradiograms of cross-linking reactions. This material is
available free of charge via the Internet at http://pubs.acs.org.
(30) Jones, J. W.; Robins, R. K. J. Am. Chem. Soc. 1963, 85, 193–201.
(31) Chang, C.-j.; DaSilva, J.; Byrn, S. R. J. Org. Chem. 1983, 48, 5151–
5160.
JA802177U
9
10306 J. AM. CHEM. SOC. VOL. 130, NO. 31, 2008