ABSTRACT
Pyrimidine/purine/purine triplexes are known to inhibit DNA polymerization. Here
we have studied the mechanisms of this inhibition by comparing the efficiency
of Vent DNA polymerase on triplex- and duplex-containing templates at different temperatures, Mg
2+
concentrations and time intervals with the thermal stability of the
corresponding structures. Our results show that triplexes can only be by-passed at temperatures where thermal denaturation initiates, while
duplexes, in contrast, are overcome at temperatures where they are quite
stable. These results show that DNA polymerase cannot untangle triplex regions
within DNA templates and seems to entirely depend on their thermal
fluctuations. The high stability of triplexes at physiological temperatures and
ambient conditions make them a barrier to polymerization.
During the last decade, triplex DNA has attracted considerable attention from a
structural viewpoint as a potential regulator of genetic processes and as a
promising therapeutic modality (reviewed in
1
). Two chemically distinct structures are known. Pyrimidine/purine/pyrimidine
(YR[middot]Y) triplexes are built from intertwined CG[middot]C
+
and TA[middot]T base triads, require cytosine protonation and are stable at acidic pH.
Pyrimidine/purine/purine (YR[middot]R) triplexes can be built from CG[middot]G, TA[middot]A and TA[middot]T triads (i.e. the third strand may not be
entirely purine) and are stable at physiological pH in the presence of divalent
cations. In both cases, the two chemically homologous strands (both pyrimidine
or both purine) are anti-parallel. Triplexes can be formed by a single DNA molecule (intramolecular
triplexes) or by different molecules (intermolecular triplexes). Intramolecular
triplexes can be formed by a single polynucleotide chain (
2
) or by a double-stranded DNA when one of its strands folds back to pair with the adjacent
duplex (H-DNA) (
3
).
Sequences that are able to form intramolecular triplexes are unusually frequent
in eukaryotic genomes and are predominantly located in their regulatory regions
(
4
). Their frequency and location has led to many hypotheses as to their role in
replication, transcription and recombination (reviewed in
1
,
5
). There is strong biochemical evidence that triplexes affect such basic genetic
processes as DNA and RNA polymerization (
6
-
15
). We have found a particularly strong effect of YR[middot]R triplexes on DNA polymerases. These triplexes, which are stable under
optimal polymerization conditions, efficiently halt different DNA polymerases (
6
,
8
).
The mechanisms of triplex-caused polymerization arrest, however, remain unclear. One could imagine
several possibilities. First, under polymerization conditions, with magnesium
present, YR[middot]R triplexes may be so much more stable than duplexes that triplex
blockage of polymerization is a simple reflection of their persistence. Second,
the kinetics of polymerase passage through triplexes may be slower than through
duplexes, simulating polymerization blockage. Finally, DNA polymerases, while
capable of dismantling duplexes, may be unable to do so with triplexes.
To distinguish between these possibilities, we made single-stranded DNA templates containing intramolecular YR[middot]R triplexes or the corresponding YR duplexes and studied the efficiency of
Vent(exo
-
) polymerase passage through these structures at different temperatures,
magnesium concentrations and time intervals. In parallel, the stabilities of
different triplex and duplex structures were determined using DNA melting
experiments.
Comparisons of the data revealed that although YR[middot]R triplexes are somewhat more stable than the corresponding duplexes ([Delta]
T
m
varies from 2 to 12oC depending on the sequence and magnesium concentration), this effect is
unlikely to account for the drastic differences in temperature, which range
from 20 to 40oC, at which polymerase traverses these structures. Projection of polymerase
passage temperatures onto melting curves for different structures shows that
the polymerase passes triplexes at temperatures where they initiate
dissociation, whereas duplexes are overcome far bellow these temperatures. We
conclude, therefore, that although DNA polymerase untangles duplex regions in
DNA templates, it lacks such an ability for triplex regions. The implications
of this finding for the biological role of triplexes and gene targeting are
discussed.
Vent(exo-) and Vent DNA polymerases were obtained from New England Biolabs, Taq
polymerase was from Gibco BRL, Stoffel polymerase was from Perkin Elmer and Pfu
polymerase was from Stratagene.
Oligodeoxyribonucleotides were synthesized on an ABI High Throughput DNA/RNA
Synthesizer Model 394 as described in the user's manual (Applied Biosystems).
They were deprotected by incubation in concentrated ammonium hydroxide for 6 h
at 65oC, followed by precipitation with 2 vol. 2 M LiClO
4
solution in acetone. The dried pellets were dissolved in TE buffer and
additionally purified by electrophoresis on a 12% denaturing polyacrylamide gel
(
16
). Upon standard elution (
16
), the oligonucleotides were ethanol precipitated and filtered through NAP-5 columns (Pharmacia).
Oligonucleotides of interest were first cloned in the pUC19 polylinker between
the
Bam
HI and
Eco
RI sites. In order to obtain single-stranded DNA,
Eco
RI-
Hin
dIII fragments from the pUC derivatives were recloned into the phagemid vector
pBluescript SK(-) (Stratagene). All plasmids were maintained in the
Escherichia coli
strain XL1-Blue (Stratagene). Supercoiled plasmid DNA was isolated using the Wizard
DNA Purification Kit (Promega). Phages were obtained as described by
Samadashwily
et al
. (
6
). Single-stranded (ss)DNA was isolated from phage particles using the QIAprep Spin
M13 Kit (Qiagen).
Experiments with Vent polymerase were carried out as follows. For annealing, 1
pmol ssDNA and 40 pmol reverse primer were incubated for 3 min at 65oC in 20 [mu]l 7.5 mM MgSO
4
, 15 mM KCl, 15 mM (NH
4
)
2
SO
4
, 30 mM Tris-HCl, pH 8.8, after which samples were slowly cooled to room temperature.
Labeling was carried out for 30 min at 25oC in 30 [mu]l 0.125 [mu]M each dGTP, dCTP and dTTP, 2 [mu]Ci [[alpha]-
32
P]dATP (3000 Ci/mmol; Amersham), 5 mM MgSO
4
, 10 mM KCl, 10 mM (NH
4
)
2
SO
4
, 20 mM Tris-HCl, pH 8.8, and 0.1% Triton X100 with 3 U polymerase. Subsequently,
samples were cooled to 0oC, diluted 10-fold and then ionic conditions were re-adjusted so that the primer extension was carried out in 7 [mu]l 10 mM KCl, 10 mM (NH
4
)
2
SO
4
, 20 mM Tris-HCl, pH 8.8, and 0.1% Triton X100, 100 [mu]M of each dNTP and 1-8 mM MgSO
4
. Temperature and time intervals of primer extension varied as described in
Results.
For Pfu polymerase, the polymerization buffer was supplemented with 100 [mu]g/ml BSA. For Taq and Stoffel enzymes, polymerization was performed in 10 mM
Tris-HCl, pH 8.8, 50 mM KCl, 100 [mu]M each dNTP and 4 mM MgCl
2
.
Radiolabeled primer extension products were separated on an 8% sequencing gel,
followed by autoradiography. The temperature at which the efficiency of
polymerization through a structure was ~50% as determined by visual inspection of autoradiograms was called the
temperature of polymerase passage (
T
pp
).
DNA melting experiments were performed using a Beckman DU 7400 spectrophotometer
equipped with the Micro
T
m
Analysis Accessory DU 600. Oligonucleotides were diluted in a buffer containing
10 mM KCl, 10 mM (NH
4
)
2
SO
4
, 0.01 mM EDTA, 20 mM Tris-HCl, pH 8.8 and 0-8 mM MgSO
4
so that the A
260
was 0.1-0.15. Samples were heated for 1.5 min at 95oC, quickly cooled to ~0oC in a dry ice-ethanol bath, incubated at 50oC for 30 min and then degassed under vacuum
for 30 min. Experimental measurements were in 1 cm quartz minicells (350 [mu]l volume). The rate of heating/cooling was 1oC/min; absorbance was measured every 0.5oC. The experimental curves were analyzed as follows. A floating
window of 21 data points was fitted by a set of third order polynomials (method
of least difference), generating a unique polynomial for each data point. The
derivatives of the polynomial at each central point,
T
, of the window was taken as the derivative of the experimental curve d
Abs
/d
T
at that point.
T
m
values were defined as the maxima of the first derivative, d
Abs/
d
T
.
To study the influence of triplex stability on DNA polymerase passage, we used
synthetic oligonucleotides that can fold into either intramolecular YR[middot]R triplexes or the corresponding YR duplexes (Fig.
1
). These oligonucleotides were cloned into the pBluescript plasmid (Stratagene)
and circular ssDNA templates containing the sequences of interest were isolated
as described (
6
).
In order to compare the efficiency of polymerization though a structure with its
thermal stability, we used Vent(exo
-
) DNA polymerase (
18
), which functions over a remarkably wide range of temperatures and magnesium
concentrations. We annealed primers to the different templates shown in Figure
1
and carried out labeling under the same conditions for all templates, followed
by primer extension for 40 min at different temperatures and magnesium
concentrations.
Polymerization efficiencies on templates with sequences capable of forming H-r5 triplexes are strikingly different from those on templates with
duplexes. Figure
2
shows the comparative results for the WT triplex- and duplex-forming sequences. The triplex-forming sequence blocks polymerization virtually completely up
to 80oC at 8 and 4 mM Mg
2+
;
the polymerase advances through it
with ~50% efficiency at 75oC at 2 mM Mg
2+
and between 60 and 65oC at 1 mM Mg
2+
. In contrast, 50% polymerase passage on the corresponding duplex-forming sequence takes place between 30 and 35oC at both 1 and 2 mM Mg
2+
.
The above results show that DNA polymerase has severe difficulties advancing
through triplexes within DNA templates while easily passing the corresponding
duplexes. However, all of these data were obtained for the same elongation time
(40 min), so it is reasonable to ask if the DNA polymerase is blocked by
triplex structures or just slowed down. At high temperature the DNA polymerase
manages to by-pass triplexes. Is this due to lower triplex stability or to an increase
in the polymerization rate?
In order to address these related questions, we began experiments in which time
courses of polymerization through different structures were studied. Figure
4
represents results for the M1 triplex and duplex at 2 mM Mg
2+
. At 32.5oC the polymerase reaches either triplex or duplex structure (from the
primer located 126 bases upstream) within the first 4 min, giving an estimate
of the polymerization rate on ssDNA at ~30 nt/min. Subsequent polymerase passage through the triplex and duplex
differs substantially. For a duplex, 50% passage requires another ~10 min (~1 nt/min). Thus, at low temperature duplex structure slows polymerase
down ~30-fold.
All the above data were obtained using Vent(exo
-
) DNA polymerase. It was essential, therefore, to determine if the observed
temperature dependence of polymerase blockage reflects a general property of
DNA polymerase-triplex interaction or is specific for Vent(exo
-
). In order to address this question, we analyzed the effect of triplexes on
several thermostable DNA polymerases, including the exo
+
version of Vent, Taq DNA polymerase, the 5'-exo
-
version of Taq (Stoffel fragment) and Pfu DNA polymerase. Note that these
enzymes differ in their processivity,
K
m
values and polymerization rate, but share strand displacement activity.
Figure
5
shows the temperature dependence of polymerization blockage by the M1 triplex
at 4 mM Mg
2+
for different enzymes. As one can see, the
T
pp
is 70oC for Vent(exo
-
), ~75oC for Vent(exo
+
), Taq and Pfu polymerases and ~82oC for the Stoffel fragment. We believe that these values are not
significantly different from one another. Thus, we conclude that the
polymerization inhibition effect reflects a general feature of polymerase-triplex interplay rather than the polymerase itself.
To determine the thermodynamic stability of different triplexes and duplexes we
synthesized the oligonucleotides shown in Figure
1
and carried out thermal denaturation experiments at different Mg
2+
concentrations. Note that our triplexes are purely intramolecular. It was
previously shown that such YR[middot]R triplexes melt cooperatively, resulting in increased stability
relative to the corresponding duplexes (
19
,
20
).
To compare the thermal denaturation data with the polymerization results,
melting experiments were performed in a buffer closely resembling the
polymerization buffer with up to 8 mM MgSO
4
. Oligonucleotides were first heated to 95oC to destroy preformed structures and then rapidly cooled to minimize
formation of intermolecular complexes.
Typical melting curves for the WT triplex and duplex are shown in Figure
6
. The top panel shows that in the absence of magnesium, i.e. when only the
duplex is formed, triplex-forming and duplex-forming oligonucleotides have virtually identical melting profiles.
The observed difference in the hyperchromicity was expected, since only two
thirds of a triplex-forming oligonucleotide is involved in duplex formation. In the presence
of 2 mM Mg
2+
, on the other hand, the difference between triplex- and duplex-forming oligonucleotides is evident (Fig.
6
, bottom panel). Although both melt cooperatively, the
T
m
for the triplex is 93oC, versus 82.5oC for the duplex.
Our data show that there are fundamental differences in the way DNA polymerase
responds to triplexes and duplexes within DNA templates. The polymerase
overcomes duplexes at low temperatures (~30oC), but for triplexes, temperatures 20-40oC higher are required. The two structures give rise to
differences in polymerization kinetics: triplexes slow polymerase more
dramatically than do duplexes. Kinetics results also show that the ability of
the polymerase to overcome triplexes at high temperatures is not due to an
increase in the polymerization rate. In fact, within a 10oC temperature range, the strength of polymerization inhibition (adjusted to
the polymerization rate at different temperatures) can change 10-fold. This indicates that there is a temperature-dependent transition of the triplex which enables DNA polymerases to
proceed.
Stabilities of different triplexes and duplexes were analyzed by thermal
denaturation experiments. They showed that whereas triplexes are more stable
than duplexes in the presence of magnesium ions, the difference is
substantially smaller than the difference in
T
pp
values. In fact, under a given set of conditions one can find a duplex that is
more stable than a triplex of a different composition, yet only the latter
blocks polymerization. The most vivid illustrations of the relation between the
thermal stability of a structure and the temperature optimum for polymerization
through it are presented in Figure
6
, where arrows show how the
T
pp
values relate to the melting curves for the WT triplex and duplex. Clearly,
triplexes are by-passed at temperatures where their thermal denaturation initiates, while
duplexes are overcome at temperatures where they are quite stable.
To explain these results, it is useful to consider two possible modes of DNA
polymerase movement through structured parts of a ssDNA template. It can either
dissociate them actively (note that the energy required can in principle be
provided by dNTP hydrolysis) or passively, by utilizing fluctuations in the
structure. In the latter case, by making fluctuations irreversible, the
polymerase would shift the equilibrium from the structured to single-stranded state. Our data lead us to conclude that DNA polymerase moves
passively through triple-helical regions in template DNA and seems to depend entirely on thermal
fluctuations within them. Since YR[middot]R triplexes are remarkably stable at physiological temperatures and
ambient conditions, they represent a barrier to polymerization.
The ability of many DNA polymerases to function on double-helical DNA by displacing the non-template DNA strand is well known (reviewed in
21
). Our results (see Fig.
6
) are consistent with the view that polymerase actively untangles duplexes,
probably by strand displacement.
What remains to be explained is the difference in the inhibitory effect of H-r5 and H-r3 triplexes of the same composition. In order for the polymerase to
proceed, it has to unwind the duplex part of the triplex for the H-r5 or a Hoogsteen strand alone for the H-r3 triplex. Our data show an ~20oC higher
T
pp
value for the H-r5 structure than for the H-r3 structure. If polymerization efficiency is limited by triplex
fluctuations, we assume that the Hoogsteen strand is the first to dissociate.
Although it seems to be true for intermolecular YR[middot]R triplexes (
19
), our thermal denaturation experiments and those of others (
19
,
20
) show that intramolecular H-r3 and H-r5 triplexes melt cooperatively with very similar
T
m
values. In order to resolve this contradiction, one should remember that
thermal denaturation experiments characterize the equilibrium dissociation of a
triplex, whereas the presence of polymerase makes this melting an inherently
non-equilibrium process. Moreover, since a polymerase approaches a triplex
from the 3'-end, fluctuations of the 3'-end are probably most determinant for this process.
This effect cannot be seen in equilibrium melting curves. For the 3'-end of the H-r3 triplex to dissociate, it is sufficient to break Hoogsteen
hydrogen bonds, but for the H-r5 triplex both the Watson-Crick and Hoogsteen hydrogen bonds must be broken. Thus, the
relative ease of polymerization through H-r3 triplexes may be easier.
The fact that at physiological temperatures and magnesium concentrations
polymerization cannot overcome YR[middot]R triplexes in DNA templates in a reasonable time may have important
biological implications. Sequences with triplex-forming potential are widely represented in eukaryotic DNAs, especially in
their non-coding areas (
4
). One can envision several possibilities for triplex formation
in vivo
. An increase in local DNA supercoiling resulting from changes in chromatin
(reviewed in
22
) or due to transcription (
23
) could promote formation of intramolecular triplexes (
24
). Transcription-induced triplexes could be additionally stabilized through their
interactions with RNA transcripts (
25
). Completely different opportunities arise from the coordination of DNA
synthesis of both daughter DNA strands (reviewed in
21
). The semi-discontinous mechanism suggests the transient occurrence of long (Okazaki
size) single-stranded portions of the lagging strand template. In the appropriate
sequence context, such single-stranded pieces could fold into triplexes (as happens for the sequences
presented in this paper) or could interact with downstream duplex DNA forming H-like structures (
5
). Obviously, this hypothesis assumes that ssDNA may fold into a structure prior
to being covered by SSB protein. It may not be that surprising, therefore, that
sequences with triplex-forming potential have been shown to inhibit DNA replication in several
eukaryotic systems (
26
-
29
).
The high efficiency of DNA polymerase inhibition we describe makes purine-rich triplex-forming oligonucleotides plausible candidates for targeting
polymerization in living cells. If a triplex-forming oligonucleotide (TFO) pairs with intracellular DNA, the
replication machinery must unravel the duplex component of the triplex, which
appears to be a difficult task for DNA polymerase. The same would likely be
true for DNA-dependent RNA polymerases. Currently the majority of efforts in anti-gene strategy employ TFO binding to different promoters or origins
in order to prevent initiation of transcription or replication (reviewed in
1
,
30
). In the light of our results, attempts to inhibit the elongation stage of
polymerization seem worth considering.
We would like to acknowledge Kevin Kaufman for his help with the computer
analysis of the thermal denaturation experiments. This work was supported by
grant MCB9405794 from the NSF to S.M.M. and an international supplement to this
grant.
REFERENCES
Return