Ci 1997 Oxford University Press
Nucleic Acids Research, 1997, Vol. 25, No. 4 883-887
Radioprobing of DNA: distribution of DNA breaks
produced by decay of 1251 incorporated into a
triplex-forming oligonucleotide correlates with
geometry of the triplex
Igor G. Panyutin* and Ronald D. Neumann
Department of Nuclear Medicine, Warren G.Magnuson Clinical Center, National Institutes of Health, Bethesda,
MD 20892, USA
Received September 26, 1996; Revised and Accepted December 12, 1996
ABSTRACT
The distribution of breaks produced in both strands of
a DNA duplex by the decay of 1251 carried by a
triplex-forming DNA oligonucleotide was studied at
single nucleotide resolution. The 1251 atom was located
in the C5 position of a single cytosine residue of an
oligonucleotide designed to form a triple helix with the
target sequence duplex. The majority of the breaks
(90%) are located within 10 bp around the decay site.
The addition of the free radical scavenger DMS0
produces an insignificant effect on the yield and
distribution of the breaks. These results suggest that
the majority of these breaks are produced by the direct
action of radiation and are not mediated by diffusible
free radicals. The frequency of breaks in the purine
strand was two times higher that in the pyrimidine
strand. This asymmetry in the yield of breaks correlates
with the geometry of this type of triplex; the C5 of the
cytosine in the third strand is closer to the sugar—
phosphate backbone of the purine strand. Moreover,
study of molecular models shows that the yield of
breaks at individual bases correlates with distance
from the 1251 decay site. We suggest the possible use
of 1251 decay as a probe for the structure of nucleic
acids and nucleoprotein connplexes.
strand of a DNA duplex produces strand breaks located within
10 bp from the decay site (4,5) with an efficiency close to 1 break/
decay (6).
We propose the use of 125I-labeled triplex-forming oligonucleotides (TFO) to deliver radionuclides to target DNA sequences so
that decay will produce sequence-specific DNA breaks. We have
shown that decay of 1251 in a pyrimidine TFO results in doublestrand breaks (DSB) in the target duplex with an efficiency close
to 1 break/decay (7). The yield of breaks was nearly equal in both
DNA strands, but the total yield of DSB was not high due to low
incorporation of radioiodine into the TFO in that particular case
(7). To increase the yield of DSB we labeled a purine-rich TFO
at the C5 positions of several cytosines (8). We used this labeled
TFO to form a triplex with a target sequence in the human HPRT
gene. After 30 days of decay, Southern blot analysis revealed
DSB in 25% of that single copy/genome sequence. However, we
estimated that the yields of breaks in the purine and pyrimidine
DNA strands were different. We suggested that yields of the
125I-induced breaks in the individual DNA strands depend on the
position of the decaying 1251 atom relative to the DNA.
Herein we present in detail the yield and distribution of 1251
decay-induced breaks in both strands of a synthetic DNA duplex
forming a triplex with an 125I-labeled TFO. The frequencies of
breaks in the DNA bases were derived from sequencing gel analysis
experiments. We relate the frequencies to the three- dimensional
structure of the triplex.
INTRODUCTION
MATERIALS AND METHODS
Decay of certain radionuclides produces a cascade of low energy
electrons, named after Pierre Auger who first discovered them in
1929(1). For example, decay of 1251 results in an emission of-20
electrons of varying energy in addition to the yphotons which are
normally detected by a y counter. Most of these Auger electrons
have energies <1 keV and ranges of less than several nanometers.
Their simultaneous action produces high density irradiation in the
vicinity of the decay site resembling that of an ct particle and
results in significant damage in adjacent molecular structures (for
reviews see 2,3). For example, decay of 1251 incorporated into one
Oligonucleotides were synthesized on an ABI-394 DNA synthesizer
(Applied Biosystems) followed by purification from a polyacrylamide gel (PAG). The purine and pyrimidine strands of the duplex
(10 pg each) were annealed by slow cooling from 100°C to room
temperature and the resulting duplex was purified from a PAG as
described (9).
The TFO was labeled with 1251 by primer extension as described
previously (7). Briefly, primer and biotinylated template oligonucleotides (10 pmol each) were annealed in 1 x Pol buffer (10 mM
Tris—HC1, pH 7.5, 10 mM MgC12, 50 mM KC1, 1 mM DDT) by
*To whom correspondence should be addressed. Tel: +1 301 496 8308; Fax: +1 301 496 0114; Email:[email protected]
884 Nucleic Acids Research, 1997, Vol. 25, No. 4
5 min incubation at 65°C followed by slow cooling to room
temperature. A primer extension reaction was carried out in the
presence of 1 mM dATP, TTP and dGTP and 100 pCi lyophilized
[5-12511o:1C1P (2200 Ci/mmol; DuPont-NEN) by 0.5 U exonucleasefree Klenow fragment of DNA polymerase I (US Biochemicals)
in a total volume of 10 p.1 at room temperature. After 10 min the
reaction was stopped by addition of 30 10 mM EDTA, the
DNA extracted with phenol:chloroform and the product duplex
purified by gel filtration through a G-50 microcolumn (Pharmacia)
equilibrated with STE buffer (50 mM Tris—HC1, pH 7.5, 1 mM
EDTA, 150 mM NaC1). At this step the amount of incorporated
[125I]dCe
was calculated as the ratio of total
radioactivity loaded
c
e
onto the column to the radioactivity that passed through.
The labeled
duplex was bound to Streptavidin Dynabeads
C
e
(Dynal) by incubation for 30 min at room temperature in STE buffer.
The Dynabeads were collected with a magnet and resuspended in
water. The duplex was denatured by heating at 80°C for 1 min and
the Dynabeads with bound template oligonucleotides were
immediately collected on ice. Supernatant containing [125I] TFO
was further purified by gel filtration into TE buffer (50 mM
Tris—HC1, pH 7.5, 1 mM EDTA) and stored at —70°C.
To forma triplex, the duplex (1 pmol) and [125I]TFO (1.5 pmol)
were mixed in lx binding buffer (50 mM Tris—HC1, pH 8, 10 mM
MgC12, 1 mM spermidine) in a total volume of 10 pl After
incubation at 37°C for 1 h, the samples were evenly divided and
to one half was added DMSO to a final concentration 1 M. Triplex
formation was analyzed by band shift assay in a 20% native PAG
using TBE containing 5 mM MgC12 as an electrode buffer (10).
The samples without DMSO were frozen at —70°C and the DMSOcontaining samples were kept at 4°C for decay accumulation.
After 30 days, 2 jiJ aliquots of the samples were diluted in 10111
2x TES (100 mM Tris—HC1, pH 8,100 mM NaC1, 2 mM EDTA)
containing 1 pmol biotinylated template oligonucleotides bound
to Streptavidin Dynabeads. The Dynabeads were removed with a
magnet and the supernatant was purified by gel filtration through
a G-50 microcolumn (Pharmacia) equilibrated with TE buffer.
The samples were 3'-end-labeled with the Klenow fragment of
DNA polymerase I (US Biochemicals) and either [cx-32P1dATP
or [ot-3211dCTP and analyzed in a 10% denaturing PAG.
Maxam—Gilbert sequencing reactions were performed as described
(11). Oligo(dT) markers of 8-32 nt (Pharmacia) were labeled by
T4 polynucleotide kinase (Gibco-BRL) with [y-32flATP. The gel
was fixed in 10% acetic acid, dried and quantitated with a BAS
1500 Bio-Imaging Analyzer (Fuji).
Molecular modeling was performed on a PowerMac computer
using MacImdad software (Molecular Application Group). The
NMR-based molecular structure of the triplex (12) was imported
from the Brookhaven Protein Database (PDB accession no. 134d).
The interatomic distances in the computer-simulated triplex
model were based on the structure proposed by Zhurkin et al. (13)
and were kindly provided by Dr Victor Zhurkin (NCI).
RESULTS
Figure 1 shows the sequences of the oligonucleotides and the
schemes of 1251 labeling and triplex formation. Although the TFO
contains both G and T residues, we will refer to this triplex as
YR.R-type, following the classification suggested by FrankKamenetskii and Mirkin (14). To incorporate [125I]dC into a
single position of the TFO we annealed the primer oligonucleotide,
which is identical to the 5'-end of the TFO, to the biotinylated
A
Primer extension
Primer
Triplex
TFO
5' -e
5 ' - ATICICTA
3' -
TA
t e
ce re
01 4C ,ee
Ce ...Ce AGAAAAAMC
Duplex
B
1234
Figure 1. (A) Scheme for labeling TFO by primer extension reaction and triplex
formation. The bases that are incorporated by DNA polymerase are shown in
italics. [125I]dC is marked by an asterisk. The template is shown attached to a
magnetic Dynabead. (B) Band shift assay of triplex formation in a 20% PAG.
Lane 1, pBR322 digestion by Mspl as length markers; lane 2, 32P-labeled
duplex; lane 3, triplex formation between unlabeled duplex and 11251]TFO; lane 4,
same sample as in the lane 3 after addition of DMSO. Arrows indicate the
position of the triplex (T), position of the duplex (D) and position of the
[125I]TFO. The lengths of some fragments in bp are shown on the left.
template (Fig. 1A). Then the 3'-end of the resulting du_plex was
extended by DNA polymerase in the presence of [5-125IldelT
and unlabeled dGTP and dTTP. Gel filtration through a Sephadex
G-50 MicroSpin column revealed that 15 i.tCi 1251 was incorporated into the duplex. Assuming that all 10 pmol of initial duplex
passed through the column, the efficiency of incorporation was
estimated as 0.7 1251 per TFO (1251 sp. act. 2.2 liCi/pmol). The
template—TFO duplex was denatured by heating, the template was
precipitated on streptavidin-modified magnetic beads (D)maheads) and removed with a magnet.
The [125I1TFO was annealed to a 50mer duplex containing a
truncated polypurine/polypyrimidine sequence from the HPR1
gene (Fig. 1) (8). The duplex has 5' overhangs; AT on the
pyrimidine strand and GC on the purine strand. Therefore, the
strands can be selectively labeled at the purine or pyrimidine
strand by DNA polymerase using combinations of [a-32P]dATP
and unlabeled dTTP or [a-32P]dCTP and unlabeled dGTP.
Triplex formation between [125ETFO and the duplex was
confirmed by band shift in a 10% native PAG (Fig. 1B). By
measuring the intensities of the bands corresponding to the free
Nucleic Acids Research, 1997, Vol. 25, No.4
885
4.D
3.S
3.D
2.S
!f"
.,
2.D
.c
1.S
.
""e
'0
>-
u
C
II
::J
tT
1.D
D.S
~
Purine strand
G GGAA GGGA
l
GGG~G
GGA GA AA GG A G GA AG GG
C C C T Tee eTC C eTC C eTC T T Tee Tee T Tee C
D.S
~~~~~m~~~~~~
Pyrimidine strand
1.D
Figure 3. Summary of break distribution on both strands of the duplex.
Frequencies are shown to the same scale for the purine and pyrimidine strands.
Open bars represent the experiment in the absence of DMSO and black bars in
the presence of DMSO. The position of labeled C is marked with an asterisk.
Figure 2. Analysis of the breaks. Autoradiograph of a 12% PAG. Lane I,
oligo(dT) 8-32 nt marker (Pharmacia); lane 2, Maxam-Gilbert G sequencing
lane of the purine strand of the duplex; lane 3, purine strand, no TFO added;
lane 4, decay accumulation at -70°C; lane 5, decay accumulation at 4°C in I
M DMSO; lane 6, pyrimidine strand, no TFO added; lane 7, decay
accumulation at -70°C; lane 8, decay accumulation at 4°C in I M DMSO; lane
9, Maxam-Gilbert C sequencing lane of the pyrimidine strand of the duplex; lane
10, 125I-labeled TFO.
[1251]lFO and the triplex we estimate that 64% of the radioactivity
was associated with the triplex band.
To accumulate a measurable number of decay-induced breaks,
the triplex samples were kept for 30 days at -70°C or at 4°C in
the presence of the free radical scavenger DMSO. Then the
triplexes were disrupted by chelating the magnesium required for
stabilization of the YRR triplexes with EDTA. The lFO was
removed from the samples by the Dynabead-bound complementary
template and precipitated with a magnet. The supernatant contained
<2% of the original radioactivity, while 98% was precipitated
with the Dynabeads. Free from [1251]TFO, duplexes were labeled
at either the purine or the pyrimidine strand and decay-induced
breaks were analyzed in a 10% denaturing PAG (Fig. 2).
The bands representing the shorter fragments corresponding to
the breaks are numbered in lanes 5 and 8 in Figure 2. Relative
intensities of the bands were measured and, after subtraction of
the intensities of the corresponding areas in the blank control
lanes 3 and 6 (background), expressed as a percentage of the total
radioactivity in the lanes, including the top band of undamaged
fragment. These values are here called frequencies of breaks (Fi,
where i is the number of the base from the labeled 3' -end) and are
presented in Figure 3 as a bar graph. The position of the breaks
relative to the sequence of the duplex was determined by comparison
with the bands corresponding to the breaks with the Maxam-Gilbert
sequencing ladders (lanes 2 and 9). Therefore, a break at a given
position means complete removal of the corresponding base, as in
the case of Maxam--Gilbert chemistry (11).
The total frequency (F) of 1251 decay-induced cleavage of the
purine and pyrimidine strands (or, in other words, the percentage
of the strands broken by 1251 decay) was calculated as the sum of
the frequencies of breaks corresponding to individual bases (F i).
The fraction of 1251 that decayed in D days can be calculated as
1_2-D/ T / 2 , where T V2 =60 days is the half-life of the isotope. Thus,
after D = 30 days 29% of the 125 1 decayed. Taking into account
that the efficiency of 1251 incorporation into lFO was 0.7 (see
above) and, therefore, at maximum F can reach only 70%, we
calculated the efficiency of strand cleavage per decay as Y =
F/(0.7 x 29). The data are summarized in Table 1.
Table I. Summary of strand cleavage after 30 days
-DMSO
+DMSO
F
y
F
y
Purine strand
17.8
0.88
16.9
0.83
Pyrimidine strand
9.0
0.44
7.9
0.39
F, total frequency of strand cleavage (%).
Y, efficiency of strand cleavage per decay of 1251.
The observed frequencies of breaks (Fi) plotted in Figure 3 do
not reflect the true probabilities of breaks/decay at the corresponding
bases. Indeed, if there is > I break/decay then only the break
closest to the 32P-Iabeled 3' -end of the strand will be detected. To
886 Nucleic Acids Research, 1997, Vol. 25, No. 4
Distances
(4
NMR
Com p.
14.9
14.9
14.1
15.7
16.3
16.5
17.7
17.5
TA
„
MI
Distances (ik)
c a ME
C G ME
Com p. NMR
T A
C G
C G
C G T*A
C G
13.7
14.5
15.7
0.06
10.0
6.3
7.4
9.8
13.4
16.7
6.1
9.1
13.4
17.4
19.9
C G
C G
A
C G In
15.9
15.8
15.6
9.6
7.8
A
AU
A
C al
C al
Ak
C al
C G1
A
T A I
I 0.12
18.9
18.9
0.24
I
I
I
0.36
Probability of break (m)
Figure 4. Calculated probabilities of breaks. Only the experiment in the
presence of DMSO is shown. Numbers show averaged distances in angstroms
from the suggested position of the iodine atom to the carbons of deoxyribose
of the corresponding base. NMR, distances calculated based on NMR data;
Comp., distances calculated based on computer-simulated triplex structure.
obtain the probabilities of breaks/decay (pi) we first calculated the
efficiencies of cleavage/decay at individual bases, Yi Fi/(0.7 x 29).
The efficiencies can be expresse(' in terms of probabilities as Yi =pi
(1 - Pi- i)(1 -pi 2). . (1 - l) , where the terms in parentheses
account for the fact that in order to observe the band corresponding
to the fragment i nt long there should be no breaks between the
1 and i - 1 positions. Reversing the above expression one can
derive the recursive expression for the probability of break/decay
at the ith position, pi = Yi/(1 - pi _ 1)(1 - Pi-2)...(i - Pl).
In practice we limited the calculation to the bands that are
numbered in Figure 2. For example, since there are no visible
bands below band 14 in lanes 4 and 5 (corresponding to A14 of
the purine strand of the duplex), we consider Y14 = P14. Then the
probability of a break at position 15 is /315 = Y15/(1 - Pi4), the
probability of a break at position 16 is p16 = Y16/(1 - p 15)(1 -P14)
and so on up to position 36. These calculations were done for both
DNA strands and the results in the form of a bar graph are
presented in Figure 4. The sum of all pi values along a strand,
which corresponds to the average number of breaks/decay/strand,
is 2.1 for the purine and only 0.5 for the pyrimidine strand. As a
consequence of this, the difference in the probabilities of breaks
between the purine and pyrimidine strands is considerably more
pronounced than the same difference in the frequencies of breaks
(compare Figs 3 and 4). The maximum of the pi distribution in the
purine strand is shifted 1 base towards the 5'-end relative to the
maximum of the Fi distribution and the iodine-containing triad.
There are no significant changes in the shape of the pi distribution
in the pyrimidine strand as compared with the corresponding Fi
distribution.
DISCUSSION
We determined for the first time the distribution of TFO-delivered
1251 decay-induced breaks in both strands of a DNA duplex. The
breaks are localized in proximity to the decay site; 90% of the
breaks are within 10 bp from the position of 1251. Decay
accumulation in the presence of the free radical scavenger DMSO
showed only a 5% reduction in the total yield of breaks in the
purine and 12% in the pyrimidine strand of the duplex. The fact
that DNA cleavage effectively occurs in the presence of DMSO
suggests that DNA cleavage was mainly induced by the direct
action of Auger electrons on DNA and bound water rather than
by diffusible free radicals produced by decay in solution. Although
the mechanism of DNA cleavage by Auger electrons is complicated
and not completely understood, it is commonly suggested that the
main targets in the DNA molecule are deoxyriboses and phosphate
groups (15-18). Several computer models have been developed
to simulate DNA breaks by Auger electrons (15-18). Calculations
based on these models predicted distributions of breaks that are
in satisfactory agreement with the experimental data for a single
DNA strand (4).
The main feature of the data presented in Figure 4 is the
significantly higher number of breaks in the purine strand versus
the pyrimidine strand. In our previous study, using a pyrimidine
TFO that formed a protonated YR.Y triplex, we estimated that the
yield of breaks was almost the same in both strands,(7). To explain
the difference between the two triplexes we turned to molecular
modeling. The TA.[1251] C triad of the YR.R triplex (19) and
CO- [125I]C+ triad of the protonated YR.Y triplex are presented in
Figure 5. The C+ in the CG4125I1C+ triad bound to the WatsonCrick pair by Hoogsteen base pairing (14). The distances from the
iodine atom to the deoxyriboses (shaded), which are considered
to be the main targets for Auger electrons, of both the purine and
pyrimidine strands are almost the same in the CG.1125I1C+ triad.
In the TA1125IK triad the third base is bound via reverse
Hoogsteen base pairing and the deoxyribose of the purine base is
considerably closer to the position of the iodine atom than one on
the pyrimidine strand. Therefore, the difference in the number of
breaks/strand reflects the geometry of the DNA triads in YR.Y
and YR.R triplexes.
Using a molecular model of the triplex based on NMR data (12)
we measured the distances from the position of the methyl group
of the central T (T22; 12) in the third strand to the carbons of
deoxyribose of the bases in both strands of the duplex. The
position of the methyl group of the T in the TA-T triad is close to
the suggested position of the iodine atom in the TA.0 triad (19;
see Fig. 5). The same distances were also calculated for the
computer-simulated model of triplex proposed in Zhurkin et al. (13).
The average distances are presented in Figure 4.
There is a clear correlation between the distances and the
probabilities of breaks along the same strand; the shorter the
distance the higher probability of a break at a given base. There
is a discrepancy in the probability of breaks and the distance if one
considers residues in different DNA strands. This difference may
be due to non-linear features of the mechanism of radiation-induced
DNA break which are yet to be understood. For example, energy
transfer along the purine strand from the bases that are very close
to the iodine atom to neighboring bases would result in a higher
overall yield of breaks in this strand.
The maxima of the break probabilities almost perfectly
coincide with the shortest distances from the suggested decay site.
Interestingly, the maxima of the breaks in both strands do not
correspond to the iodine-containing triad. They are shifted 5' for
the purine and 3' for the pyrimidine strand. The same phenomenon
happens with the distances calculated for two completely
unrelated triplex models.
Nucleic Acids Research, 1997, Vol. 25, No. 4
YR•R Triplex
887
transparent to Auger electrons and, therefore, the probability of
Auger electron-induced breaks depends mostly on the distance
from the decaying atom. Although, due to the stochastic nature of
the Auger process, there is no simple analytical relationship between
distance and break probability, there are computer simulation
methods that are able, in principle, to predict break probability on
the basis of DNA strucrure (15-18). We believe that by using this
approach unique information could be obtained in solution and
even in vivo about the structures of nucleic acids and nucleoprotein
complexes.
ACKNOWLEDGEMENTS
We are thankful to Dr V.Zhurkin for computer modeling of the
triplex, Dr V.Malkov for stimulating discussions and Drs S.Mirkin
and H.Nikjoo for critical comments on the manuscript.
REFERENCES
Figure 5. TA.T (A) and TA.0 (B) triads of the YR.R triplex and the CG-C+ triad
(C) of the YR.Y triplex. Positions of the iodine atom and the sugars are shaded.
The observed correlation between the probability of breaks and
distance raises the very intriguing possibility of assessing relative
distances between DNA strands and even the bases in different
DNA structures using an Auger emitter as the probe. This method
of radioprobing DNA structure is distinct from the popular free
radical-mediated affinity cleavage technique (20). Free radicals
attack the closest residues accessible by diffusion. The reactivity
of free radicals is affected by the presence of DNA-bound
molecules and conditions in solution. In contrast, DNA is almost
1 Auger,P. (1925) J. Physique, 6S, 205.
2 Sastry,K.S. (1992) Med. Phys., 19, 1361-1370.
3 Adelstein,S.J. (1993)Am. J. Roentgenol., 160, 707-713.
4 Martin,R.F. and Haseltine,W.A. (1981) Science, 213, 896-898.
5 England,P.R. and Murray,V. (1993) Antisense Res. Dev., 3, 219-224.
6 Krisch,R.E. and Sauri,C.J. (1975) Int. J. Radiat. Biol. Relat. Studies Phys.
Chem. Med., 27, 553-560.
7 Panyutin,I.G. and Neumann,R.D. (1994) Nucleic Acids Res., 22,
4979-4982.
8 Panyutin,I.G. and Neumann,R.D. (1996) Acta Oncol., 35, 817-823.
9 Panyutin,I.G. and Hsieh,P. (1993) J. Mol. Biol., 230, 413-424.
10 Cooney,M., Czemuszewicz,G., Postel,E.H., Flint,S.J. and Hogan,M.E.
(1988) Science, 241, 456-459.
11 Maxam,A.M. and Gilbert,W. (1980) Methods Enzymol., 65, 499-525.
12 Radhakrishnan,I. and Patel,D.J. (1993)J. Am. Chem. Soc., 115, 1615.
13 Zhurkin,V.B., Raghunathan,G., Ulyanov,N.B., Camerini-Otero,R.D. and
Jernigan,R.L. (1994) J. Mol. Biol., 239, 181-200.
14 Frank-Kamenetskii,M.D. and Mirkin,S.M. (1995) Annu. Rev. Biochem.,
64, 65-95.
15 Charlton,D.E. and Humm,J.L. (1988) Int. J. Radiat. Biol. Relat. Studies
Phys. Chem. Med., 53, 353-365.
16 Pomplun,E. (1991) Int. J. Radiat. Biol., 59, 625-642.
17 Nikjoo,H., O'Neill,P., Terrissol,M. and Goodhead,D.T. (1994) Int. J.
Radiat. Biol., 66, 453-457.
18 Terrissol,M. (1994) Int. J. Radiat. Biol., 66, 447-451.
19 Beal,P.A. and Dervan,P.B. (1992) Nucleic Acids Res., 20, 2773-2776.
20 Dervan,P.B. (1991) Methods Enzymol., 208, 497-515.