Alternate strand recognition of double-helical DNA
by (T,G)-containing oligonucleotides in the presence of a triple helix-specific ligand
Alternate strand recognition of double-helical DNA by (T,G)-containing oligonucleotides in the presence of a triple helix-specific ligand
Thérèse
de Bizemont
,
Guy
Duval-Valentin
,
Jian-Sheng
Sun
,
Emile
Bisagni
1
,
Thérèse
Garestier
1
and
Claude
Hélène*
Laboratoire de Biophysique, CNRS URA 481, INSERM U 201, Muséum National d'Histoire Naturelle, 43 rue Cuvier, 75231
Paris
Cedex 05,
France
and
1
Laboratoire de Synthèse Organique, CNRS UA 1387, Institut Curie-Biologie, Bâtiment 110, 91405
Orsay
,
France
Received November 1, 1995;
Revised and Accepted January 26, 1996
ABSTRACT
Triple helix formation requires a polypurine[middot]polypyrimidine sequence in the target DNA. Recent works have shown that
this constraint can be circumvented by using alternate strand triplex-forming oligonucleotides. We have previously demonstrated that (T,G)-containing triplex-forming oligonucleotides may adopt a parallel or an
antiparallel orientation with respect to an oligopurine target, depending upon
the sequence and, in particular, upon the number of 5
'
-GpT-3
'
and 5
'
-TpG-3
'
steps [Sun
et al.
(1991)
C. R. Acad. Sci
. Paris Ser III, 313, 585-590]. A single (T,G)-containing oligonucleotide can therefore interact with two
oligopurine stretches which alternate on the two strands of the target DNA. The
(T,G) switch oligonucleotide contains a 5
'
-part targeted to one of the oligopurine sequences in a parallel
orientation followed by a 3
'
-part that adopts an antiparallel orientation with respect to the second
oligopurine sequence. We show that a limitation to the stability of such a
triplex may arise from the instability of the antiparallel part, composed of
reverse-Hoogsteen C.GxG and T.AxT base triplets. Using DNase I footprinting and
ultraviolet absorption experiments, we report that a benzo[
e
]pyridoindole derivative [(3-methoxy-7
H
-8-methyl-11-[(3
'
-amino-propyl) amino] benzo[
e
]pyrido [4,3-
b
]indole (BePI)], a drug interacting more tightly with a triplex than with a
duplex DNA, strongly stabilizes triplexes with reverse-Hoogsteen C.GxG and T.AxT triplets thus allowing a stabilization of the
triplex-forming switch (T,G) oligonucleotide on alternating oligopurine-oligopyrimidine 5
'-
(Pu)
14
(Py)
14
-
3
'
duplex sequences. These results lead to an extension of the range of
oligonucleotide sequences for alternate strand recognition of duplex DNA.
INTRODUCTION
Oligonucleotides represent a class of ligands recognizing the major groove of
double-stranded oligopurine-oligopyrimidine tracts in DNA by triplex formation (
1
,
2
). Triple helix-forming oligonucleotides can artificially control gene expression by
competing with regulatory proteins, thereby inhibiting gene transcription both
in vitro
and in cell cultures. In addition they can be used as tools in molecular
biology, e.g. for gene mapping or site-directed mutagenesis (for review see
3
,
4
).
Four classes of base triplets have been described with unmodified
oligonucleotides as third strands (C.GxC+, C.GxG, T.AxT, T.AxA) (
5
). (X.YxZ base triplets: X, Y and Z are the nucleotides in the first, second and
third strand of the triplex, respectively. `x' represents Hoogsteen or reverse-Hoogsteen hydrogen bonds, `.' indicates Watson-Crick interactions.) Base triplets involving T, C or G in the third
strand may adopt two configurations; one of them corresponds to the hydrogen
bonding scheme in co-crystals of A and T derivatives referred to as the Hoogsteen configuration
(
6
). The other one involves a 180o rotation of the base in the third strand and will be referred to as
reverse-Hoogsteen configuration. Adenine residues in the third strand adopt a
single configuration for the formation of two hydrogen bonds: the resulting
T.AxA base triplet belongs to the reverse-Hoogsteen category (
7
). Whereas Hoogsteen T.AxT and C.G.xC+ base triplets are isomorphous, the other
triplets are not (Fig.
1
).
MATERIALS AND METHODS
DNA and chemical ligand
The benzo[
e
]pyridoindole derivative [3-methoxy-7
H
-8-methyl- 11-[(3'-amino-propyl) amino] benzo[e]pyrido [4,3-
b
]indole (BePI), see Fig.
2
] was synthetized as described by Nguyen
et al.
(
22
).
Oligodeoxyribonucleotides were synthesized on an automatic synthesizer by the
Institut Pasteur (Paris) using the phosphoramidite method. Full length
oligonucleotides were separated from shorter incomplete chains by
electrophoresis on denaturing 20% acrylamide gels and recovered from mashed gel
by diffusion in water. Oligonucleotide concentrations were determined by UV-absorption spectroscopy at 260 nm using a Kontron Uvikon 860
spectrophotometer at 20oC. One of the strands of the 44 base pair (bp) DNA fragment used in this
study (44B2, see Fig.
3
) was labelled at the 5'-end with [[gamma]
32
-P]ATP (Amersham) and polynucleotide kinase (BioLabs).
Ultra-violet absorption
Thermal denaturation profiles were recorded with a Kontron Uvikon 820
spectrophotometer with 1 [mu]M of each strand in 10 mM sodium cacodylate pH 7.2, 100 mM NaCl, 5 mM MgCl
2
and 0.5 mM spermine. The BePI derivative was added at concentrations indicated
in the figure legends. Thermostatic control of the sample cell holder was by
circulating liquid (70% water/30% glycerol). The temperature of the bath was
increased or decreased at a rate of 0.1oC/min, thus allowing complete thermal equilibrium of the cells between two
measurements (separated by ~10 min). All melting curves were reversible, giving the same profile
independently of whether the temperature was raised from 0 to 60oC (or 85oC) or decreased from 60 (or 85oC) to 0oC. The half-dissociation temperature of a third strand
(temperature at which 50% of the triplex are dissociated into duplex target and
free oligonucleotide) was estimated directly from melting curves or from
derivative curves.
DNase I reactions
DNase I footprinting studies were carried out as described by Duval-Valentin
et al
. (
23
), with the following modifications. The 44 bp 44B1.44B2 duplex DNA fragment (20
nM) labelled on one strand (44B2) at the 5'-end was incubated overnight with the third strand (20 [mu]M) at various temperatures in a 10 ml solution containing 20 mM
Tris-HCl pH 7.2, 50 mM NaCl, 5 mM MgCl
2
, 0.5 mM spermine and 5 mM CaCl
2
. In some experiments, the BePI derivative was added at concentrations described
in the figure legends. DNase I (0.2 [mu]g/ml final) was added and digestion was stopped after different periods of
time (depending on the temperature) by adding 1.5 [mu]l of a mixture containing 100 mM EDTA and 10% SDS. After two ethanol
precipitations, the digestion products were analyzed on a 20% polyacrylamide/7
M urea gel and autoradiographed with Fuji X-ray films at -80oC.
Gel retardation assays
Double-stranded target sequences radioactively labelled at one 5'-end (20 nM) were put in the presence of unlabelled
oligonucleotides (20 [mu]M) in a 10 [mu]l solution containing 50 mM HEPES pH 7.2, 10 mM MgCl
2
, 0.5 [mu]M BePI. After overnight incubation at a precise temperature, 2 [mu]l non-denaturing loading buffer (50% sucrose, 5 [mu]g/[mu]l tRNA) was added and samples were loaded on a 10% native
polyacrylamide gel running under the same conditions (buffer, temperature) as
the incubation gels (but without BePI). The gel was exposed to a phosphorimager
screen.
RESULTS
Figure
3
shows the sequences of oligonucleotides used in the present study. We call Box
I, the 14 bp Watson-Crick duplex with three 5'-GpA-3' steps and two 5'-ApG-3' steps and Box II,
the sequence d(A
10
G
4
)[middot]d(T
10
C
4
) with one 5'-ApG-3' step only. These two contiguous boxes are flanked by
two sequences rich in G[middot]C base pairs which stabilize the ends of the duplex.
The (T,G) oligonucleotides are designed to bind boxes I and II in a parallel orientation (oligonucleotides 14P1 and 14P2) or in
an antiparallel orientation (oligonucleotides 14A1 and 14A2) with respect of
the oligopurine target strand. The switch (T,G) oligonucleotide called 28SW is
synthesized to bind both boxes I and II by hybridization with 5'-G
2
A
3
GA
4
GA
3
-3' on one strand in an antiparallel orientation and with 5'-A
10
G
4
-3' on the other strand in a parallel orientation. The 28SW
oligonucleotide is composed of the 14A1 and the 14P2 sequences (14A1-SW and 14P2-SW, respectively) linked together by a phosphodiester linkage.
Triple helix recognition of boxes I and II
DNase I footprinting experiments were carried out to determine whether triplex
formation occurred. We have previously shown that box II was protected by 14P2 against DNase I digestion at 4oC (
18
). No footprint was detected with 14A1, 14P1 or 14A2 at the same temperature.
The `switch' oligonucleotide 28SW protected both boxes against DNase I attack
at 4oC. It should be noted that runs of (dA)[middot](dT) present in Box II are cut poorly by DNase I which can be
accounted for by their narrow minor groove (
24
).
Table 1
.
DNase I footprinting experiments performed on the double-stranded 44 bp DNA fragment with different oligonucleotides as third
strands: (
A
) without BePI; (
B
) with 0.5 [mu]M BePI
A
Third strand
Incubation temperature (oC)
4
10
15
20
25
30
35
40
14A1, 14P1, 14A2
-
-
-
-
-
-
-
-
14P2
+++
+++
++
++
++
+
+
+-
14P2-SW
+++
+++
++
+
+
+-
-
-
14A1-SW
+
-
-
-
-
-
-
-
B
Third strand
BePI 0.5 [mu]M
Incubation temperature (oC)
20
30
40
14A1
+
-
-
14P1
-
-
-
14A2
-
-
-
14P2
++
+
+-
14P2-SW
++
++
+-
14A1-SW
++
+
+-
The tables summarize DNase I protection experiments at various temperatures. The
following symbols were used to evaluate the percentage of duplex protection by
the third strand against DNase I attack: -, 0-25%; +-, 26-45%; +, 46-55%; ++, 56-85%; +++, 86-100%. `ND', non-determined. 14P2-SW and 14A1-SW indicate
the two portions of the switch 28mer (28SW) (see Fig. 3).
The oligonucleotide 14P2 (5'-T
10
G
4
-3') adopts a parallel orientation relative to the polypurine sequence
5'-A
10
G
4
-3'. The switch oligonucleotide 28SW interacts with box II in a parallel orientation and with box I in an antiparallel manner. The temperature of incubation of
the double helix in the presence of 20 [mu]M 14P2 or 28SW was increased in order to determine the relative stability of
the triplexes which was assayed by DNase I footprinting experiments (Table
1
A). The decreasing order of relative stability as a function of temperature was:
14P2 > 14P2-SW >> 14A1-SW > 14A1, 14P1, 14A2 (no binding was detected with the last three
oligonucleotides). The results show that there is no full protection of both
boxes against DNase I cleavage with the 28SW oligonucleotide at temperatures >4oC. The 14P2 portion of the switch oligonucleotide (14P2-SW) was destabilized as compared to isolated 14P2. The lack of
stability of the switch triplex formed with 28SW appears to be due-at least in part-to a poor affinity of 14A1-SW for its duplex target. 14A1-SW binding to box I does not give a clear footprint at temperatures
>4oC, in contrast to 14P2-SW which protects box II from DNase I digestion up to 30oC. Therefore the binding of the 14A1-SW sequence should be increased if one wants to stabilize
the switch triplex. We tested BePI derivatives as triple helix-specific ligands. They were described previously as specific ligands for
stretches of T.AxT base triplets in the Hoogsteen configuration (
21
).
BePI effect on triple helix formation
We have previously shown that BePI derivatives (Fig.
2
) intercalate preferentially between contiguous T.AxT triplets allowing triplex
stabilization with a third strand containing thymine and cytosine residues (
21
). The effect of BePI was studied to determine whether it could stabilize triple
helices formed by oligonucleotides containing thymine and guanine residues.
In order to study the effect of BePI on the triplexes formed between box I and 14A1 or 14P1, DNase I footprinting experiments were performed on the
duplex target incubated at 4oC with the oligonucleotides in the presence of increasing BePI
concentrations. Up to 0.1 [mu]M BePI, no triple helix was detected on box I. In the presence of 0.3 [mu]M BePI, 14A1 was bound to its target but not 14P1, whereas both 14A1 and
14P1 formed complexes at 1 [mu]M BePI. The ligand effect on triplex stability with box I could also be observed by UV absorption experiments (Table
2
). Temperatures of half dissociation (
T
m
) of the third strand 14A1 or 14P1 were obtained by absorption spectrophotometry after subtraction of the duplex dissociation curve from the triplex curves
recorded under the same conditions. Without BePI, no transition corresponding
to dissociation of the (T,G) third strand was detected. In the presence of 3 or
10 [mu]M BePI, an hyperchromism was associated with the dissociation of the third
strand from the duplex.
T
m
values were calculated from these curves or from the corresponding derivatives:
T
m
(14A1) = 34oC and
T
m
(14P1) = 26oC with 3 [mu]M BePI, and
T
m
(14A1) = 44oC and
T
m
(14P1) = 36oC with 10 [mu]M BePI (Table
2
). The triplex formed between box I and the antiparallel 14mer (14A1) was therefore more stable ([Delta]
T
m
= 8oC) than the triple helix of box I with 14P1. Additional experiments were carried out to check whether
BePI could stabilize the triplexes with 14A2 or 14P2 on box II. DNase I footprinting experiments showed that with 0.5 [mu]M BePI, no stabilizing effect of the triple-helix specific ligand on the formation of triplexes involving 14A2 or
14P2 could be detected (Table
1
). We used thermal denaturation experiments to monitor triplex formation between
the 44 bp target 44B1.44B2 and 14A2 or 14P2 in the absence or in the presence
of BePI. Absorbance at 260 nm was recorded versus temperature. No transition
could be detected with 14A2 or 14P2 without BePI. In the presence of BePI
derivatives (3 or 10 [mu]M) an hyperchromism associated with the dissociation of the third strand
from the duplex was detected on melting curves.
T
m
values of 14A2 and 14P2 were calculated from these curves or from the
corresponding derivatives (Table
2
): with 3 [mu]M BePI,
T
m
(14A2) = 35oC and
T
m
(14P2) = 45oC; with 10 [mu]M BePI,
T
m
(14A2) = 43oC and
T
m
(14P2) = 53oC. These results show that the stability corresponding to the parallel
orientation (14P2) is favoured over that of the antiparallel orientation (14A2)
even in the presence of BePI for a triple helix containing a single 5'-TpG-3' step in the third strand.
.
T
m
values (oC) obtained at 260 nm with different third strands (1 [mu]M of third strand and duplex) in the presence of two concentrations of
BePI
Third strand
BePI ([mu]M)
3
10
14A1
34
44
14P1
26
36
14A1b
46
ND
14P1b
46
ND
14A2
35
43
14P2
45
53
ND, non-determined.
Sequence effect
The high number of (G/T) steps (5'-GpT-3' + 5'-TpG-3' steps) in 14A1 and 14P1
could have a destabilizing effect on triplex formation as suggested from
molecular modeling and energy minimization (
18
). Box I was modified by decreasing the number of 5'-ApG-3' and 5'-GpA-3' steps from five to three
while keeping the same base composition (four guanine and 10 adenine residues).
Triplex formation was studied on a 26 bp duplex containing this modified box I, called box Ib (Fig.
4
A). The oligonucleotides 14A1b and 14P1b could recognize the 26 bp duplex in an
antiparallel and parallel orientation, respectively (Fig.
4
A). No transition corresponding to a triplex-to-duplex conversion was observed in the thermal dissociation curves in
the absence of BePI. Experiments performed in the presence of 3 [mu]M BePI showed superimposable thermal dissociation curves (Fig.
4
B) for 14A1b and 14P1b with a
T
m
estimated to be ~46oC, i.e. 12 and 20oC higher than that of 14A1 and 14P1, respectively. In the
presence of 3 [mu]M BePI, the stability of the triple helices was the same for the parallel
and antiparallel orientations with three (G/T) steps whereas the antiparallel
orientation was favoured over the parallel one with a higher number of such
steps (five) for a total number of 10 T.AxT and five C.GxG base triplets. In addition the lower
number of steps increased the stability of both triple helices with a larger
effect on the parallel orientation. These results are consistent with previous
calculations and experiments which suggested that the energy penalty associated
with the non-isomorphism of T.AxT and G.GxG base triplets was larger in the Hoogsteen
than in the reverse-Hoogsteen configuration (Fig.
1
). However the length of T and G stretches also plays an important role in
determining both triplex stability and BePI binding.
Sequence specificity
In order to test the sequence specificity of the switch oligonucleotide 28SW with the specific target (44B1.44B2) containing boxes I and II and with a target containing one mismatch either in box I (box Im) or in box II (box IIm) (Fig.
6
A), gel retardation assays were performed. At 10oC, pH 7.2 and with 0.5 [mu]M BePI, the comparison of duplex and triplex populations shows that
28SW binds more efficiently to the DNA target containing alternating
oligopurine sequences in boxes I and II, than to DNA targets containing one central A[middot]T base pair inversion either in box I (box Im) or in box II (box IIm) (Fig.
6
B).
DISCUSSION
Pyrimidine-rich oligonucleotides bind to double-helical DNA in a parallel orientation with respect to the
oligopurine strand of a double-stranded oligopurine-oligopyrimidine sequence. Cytosine residues need to be protonated,
and the affinity of a pyrimidine third strand decreases at physiological pH.
The use of G instead of C allows binding of the third strand in a pH-independent manner. Triplex formation with a (T,G) oligonucleotide
requires relatively high Mg
2+
concentrations (5-20 mM).
A recent study concluded that at low guanine content (from 22 to 50%) triplex
formation could not be detected with (T,G)-containing oligonucleotides whatever the orientation of the third strand (
25
). The 14P2 oligonucleotide used in our study contains 10 thymines and four
guanines (%G = 28) and is able to bind its target at 30oC without BePI. Therefore a high G content is not a prerequisite for
triplex formation with a (T,G)-containing third strand. In addition, an oligonucleotide containing many
adjacent guanines is able to self-associate to form quadruplex structures, thus decreasing the concentration
of free (T,G) oligonucleotides available for triplex formation (
26
).
One of the important parameters that must be taken into account in designing
oligonucleotides directed to the major groove of double-stranded DNA is the isomorphism of base triplets (Fig.
1
). When the C1' atoms of Watson-Crick base pairs are fixed, the position of the C1' atom in the third strand depends on the base triplet. Both
T.AxT and C.GxC+ triplets are isomorphous in the Hoogsteen configuration.
Despite the backbone distortion that results from the non-isomorphism of other base triplets, oligonucleotides differing from those
containing only pyrimidines are able to form triple helices with oligopurine.oligopyrimidine sequences of DNA. For example reports have shown that
oligonucleotides containing either G and T (
25
-
30
), G and A (
28
,
31
,
32
), or G, T and C (
33
) form triple helices.
With all nucleotides in the anti conformation, Hoogsteen hydrogen bond formation
leads to a parallel orientation of the third strand with respect to the
oligopurine target strand. Reverse-Hoogsteen hydrogen bonding corresponds to an antiparallel orientation.
Oligopyrimidines bind in a parallel orientation with respect to the polypurine
strand of target sequences. Oligopurines bind in an antiparallel orientation.
For oligonucleotides containing G and T, the orientation depends on the base
sequence. Theoretical calculations suggested that the intrinsic stability of
T.AxT and C.GxG base triplets favours the Hoogsteen mode of hydrogen bonding.
However the energy penalty associated with the distortion of the backbone
imposed by the non-isomorphism of these two triplets is less important for the reverse-Hoogsteen configuration (see Fig.
1
). Therefore it is expected that the orientation of a (T,G) oligonucleotide
depends on the number of 5'-GpT-3' and 5'-TpG-3' steps in the third strand
[called (G/T) steps in the following discussion] and on the length of the G and
T tracts. Here we have shown that the orientation of a third strand containing
10 T and four G shifts from a parallel to an antiparallel orientation when the
number of (G/T) steps increases. Footprinting experiments showed that a 14mer
containing five (G/T) steps (14A1) was bound to the purine target in an
antiparallel orientation. In the presence of BePI the antiparallel orientation
was still more stable than the parallel one. When the number of (G/T) steps
decreased from five to one with the same base composition, the orientation
became parallel (oligonucleotide 14P2) as previously observed with another
oligonucleotide (5'-T
4
CT
4
G
6
-3') targeted to the polypurine tract of HIV (
33
). With three (G/T) steps (oligonucleotides 14A1b or 14P1b), the
T
m
values for third strand dissociation were identical whatever their orientations
in the presence of 3 [mu]M BePI [
T
m
(14A1b) =
T
m
(14P1b) = 46oC]. It should be noted that a triplex formed with an oligonucleotide
containing five (G/T) steps (14A1, antiparallel) was less stable than that
formed with one 5'-TpG-3' step (14P2, parallel). Moreover, when the number of
(G/T) steps was decreased from five to three, the triplex stability was
increased in the presence of 3 [mu]M BePI [
T
m
(14A1b) =
T
m
(14P1b) = 46oC] > [
T
m
(14A1) = 34oC,
T
m
(14P1) = 26oC)].
A comparison of the stability of triplexes with (T,G) third strands using UV
spectroscopy is difficult in the absence of BePI because no transition could be
detected in melting curves with a (T,G)-containing third strand without triplex-specific ligand. For example, even though a DNase I footprint on
44B1.44B2 was detected with 14P2, showing that a triple helix was formed under
these conditions without BePI, no hyperchromism was detected on the melting
curves. Oligonucleotides containing stretches of guanosine residues are known
to form quadruplexes. Therefore self-association could compete with triplex formation (
26
). However this competition should be detected both by footprinting and UV
melting. The hyperchromism associated with thermal dissociation is determined
to a great extent by stacking interactions. A strong stacking interaction in
single-stranded oligonucleotides containing G and T (or G and A) might not give
rise to an additional hypochromism when the triplex is formed. This could
explain the absence of transition in UV melting curves even when footprinting
experiments reveal triplex formation. The absence of hyperchromism has
previously been documented in the case of triplexes involving 8-oxoadenine (
34
).
In the presence of BePI, a transition appears in the thermal dissociation
curves, allowing us to determine the
T
m
values of (T,G)-containing third strands. This could be useful for the comparison of the
spectroscopic behavior of various (T,G) oligonucleotides which give no
hyperchromism but the BePI effect on stability should be evaluated for each
(T,G) oligonucleotide tested.
We previously proposed a correlation between the orientation of a (T,G)-containing third strand and the number of (G/T) steps (
18
). Here we have shown that a 14mer 5'-T
10
G
4
-3' with one 5'-TpG-3' step binds in a parallel orientation. A
systematic study concerning the orientation of (T,G)-containing third strands of various lengths, sequences and base
compositions will be useful to fully understand the hybridization properties of
(T,G)-containing oligonucleotides. Nevertheless, the results presented in Table
2
clearly show that both the number of (G/T) steps and the distribution of T and
G tracts are important to determine triplex stability
in the presence of BePI
. For a given sequence the more stable complex (parallel or antiparallel) is
predicted on the basis of the number of (G/T) steps. For example T
3
GT
4
GT
3
G
2
with five (G/T) steps is more stable with an antiparallel orientation; in
contrast T
10
G
4
with a single (G/T) step is more stable in the parallel orientation; the
oligonucleotide T
5
G
2
T
5
G
2
with three (G/T) steps has an intermediate behavior and binds equally well in both orientations. However it is difficult to predict the respective
intrinsic stability of two oligonucleotides in the same orientation. Not only
the number of (G/T) steps but also the lengths of G and T tracts play an
important role. For example in the presence of 3 [mu]M BePI, T
3
GT
4
GT
3
G
2
and T
10
G
4
have the same stability in the antiparallel orientation despite the great
difference in number of (G/T) steps. Similarly T
10
G
4
and T
5
G
2
T
5
G
2
have similar stability in the parallel orientation (Table
2
). Clearly the intrinsic triplex stability is the resultant of several factors:
the cooperativity of triplet stacking within T.AxT or C.GxG sequences, the
energy penalty associated with backbone distortion at each (G/T) step (in both
the third strand and the double helix) and the stacking interaction between a
T.AxT and a neighbouring C.GxG triplet (in both 5' -> 3' and 3' -> 5' orientations).
In order to recognize oligopurine sequences alternating on the two strands of
DNA, one can take advantage of the different orientations adopted by (T,G)
oligonucleotides according to their sequence. Footprinting experiments showed
that a switch oligonucleotide of 28 nucleotides (28SW) was able to bind an
alternating target duplex 5'-(Pu)
14
(Py)
14
-3' in the absence of any ligand (
18
). The switch 28mer oligonucleotide recognized box II with a parallel orientation via Hoogsteen C.GxG and T.AxT base triplets
and box I with an antiparallel orientation via reverse Hoogsteen C.GxG and T.AxT base
triplets. Although 28SW protected boxes I and II against DNase I digestion, the N7 of guanine residues in box I were not protected against DMS attack in contrast to guanine residues G19
to G22 in box II which were partially protected (results not shown). In box I, most of the guanine residues are located between two thymine residues. As
shown by NMR experiments a distortion occurs at 5'-GpT-3' and 5'-TpG-3' steps due to the non-isomorphism of base-triplets (see
Fig.
1
) (
35
). This could explain the reactivity of these guanine residues to DMS reaction.
In the presence of BePI, DNase I footprinting experiments performed at 30oC showed that the switch oligonucleotide 28SW formed a triplex that was
more stable than that obtained with separated 14A1 or 14P2 oligonucleotides.
Moreover, whereas no footprint was detected on box I with the sequence 14A1 alone , it was possible to detect a footprint on box I when this sequence was linked to 14P2 (28SW oligonucleotide). Parallel
binding of the 5'-T
10
G
4
-3' portion of 28SW induced the antiparallel fixation of the 14A1
portion of 28SW. In contrast, in the absence of BePI the triple helix formed by
28SW was less stable than that formed by 14P2 alone. Therefore, when the 3'-half of the 28SW oligonucleotide (14A1-SW) was not bound, this negatively-charged `tail' attached to 14P2-SW destabilized the triplex formed with box II.
BePI derivatives strongly stabilize triplexes with a (T,G) third strand in a
concentration-dependent manner. Previous results obtained with (T,C) oligonucleotides
have shown that BePI stabilization was observed with triplexes rich in adjacent
Hoogsteen T.AxT base triplets suggesting that BePI is preferentially intercalated between T.AxT base triplets (
21
). According to these previous results, the triplex with T
10
G
4
oligonucleotide in the Hoogsteen configuration was expected to be stabilized by
BePI derivatives as experimentally observed (Table
2
). Here we have shown that BePI can also stabilize triple helices involving
T.AxT and C.GxG base triplets in the reverse Hoogsteen configuration. Triplexes
involving five (14A1), three (14A1b) or one (14A2) (G/T) steps in the third
strand were stabilized.
As shown by DNase I footprinting experiments at increasing temperatures, the
binding of a triplex-forming switch (T,G) oligonucleotide to an alternating oligopurine-oligopyrimidine sequence is stabilized by 25oC upon binding of BePI derivatives.
Gel retardation assay showed that 28SW can discriminate, even in the presence of
BePI, between duplex DNA sequences that differ by one A[middot]T base pair inversion. This is in agreement with previous studies
showing that replacement of a single base pair in the center of an oligopurine.oligopyrimidine sequence is highly detrimental to triplex formation (
36
,
37
).
CONCLUSION
We have shown that natural oligonucleotides containing thymine and guanine
residues can recognize, in a pH-independent manner, oligopurine sequences that alternate on the two
strands of the duplex. This `switch' triplex can be stabilized by BePI
derivatives. The experimental results presented here also show that alternate-strand triple-helix formation with (T,G)-containing oligonucleotides requires appropriate target
sequences with defined numbers of 5'-ApG-3'/5'-GpA-3' steps in each of the
oligopurine target sequence and an adequate length of G and T stretches in the
third strand. A systematic study concerning the orientation of (T,G)-containing third strands of various lengths, sequences and base
composition will be useful to understand the hybridization properties of (T,G)-containing oligonucleotides and therefore to extend the number of
oligonucleotide sequences that can be used to target duplex DNA at specific
sites where oligopurine sequences alternate on the two strands of the double
helix.
ACKNOWLEDGEMENTS
Thérèse de Bizemont was supported by grants from the Ministère de l'Enseignement Supérieur et de la Recherche. We thank Olavio
Delgado and Loïc Perrouault for the photography.
REFERENCES
1 Le Doan,T., Perrouault,L., Chassignol,M., Thuong,N.T. and Hélène,C. (1987) Nucleic Acids Res., 15, 8643-8659.
2 Moser,H.E. and Dervan,P.B. (1987) Science, 238, 645-650.
