ABSTRACT
We have examined the effect of a naphthylquinoline triplex-binding ligand on the formation of intermolecular triplexes on DNA
fragments containing the target sites A
6
G
6
[middot]C
6
T
6
and G
6
A
6
[middot]T
6
C
6
. The ligand enhances the binding of T
6
C
2
, but not T
2
C
6
, to A
6
G
6
[middot]C
6
T
6
suggesting that it has a greater effect on T[middot]AT than C
+
[middot]GC triplets. The complex with T
6
C
2
is only stable below pH 6.0, confirming the requirement for protonation of the third strand cytosines. Antiparallel triplexes with GT-containing oligonucleotides are also stabilised by the ligand. The complex between G
5
T
5
and A
6
G
6
[middot]C
6
T
6
is stabilised by lower ligand concentrations than that between T
5
G
5
and G
6
A
6
[middot]C
6
T
6
. The ligand does not promote the interaction with GT-containing oligonucleotides which have been designed to bind in a parallel
orientation. Although the formation of antiparallel triplexes is pH
independent, we find that the ligand has a greater stabilising effect at lower
pH, suggesting that the active species is protonated. The ligand does not
promote the binding of antiparallel GA-containing oligonucleotides at pH 7.5 but induces the interaction between
A
5
G
5
and G
6
A
6
[middot]T
6
C
6
at pH 5.5. Ethidium bromide does not promote the formation of any of these
triplexes and destabilises the interaction of acridine-linked pyrimidine-containing third strands with these target sites.
The formation of DNA triple helices was first described in 1957 (
1
) and is an important means of achieving sequence specific recognition of duplex
DNA by synthetic oligonucleotides (
2
-
4
). Sequence specificity is achieved by the formation of hydrogen bonds between
bases in the third strand and major groove substituents on the duplex purine
strand. Two different types of DNA triple helix have been described which
differ in the orientation of the third strand with respect to the purine strand
of the duplex. The parallel motif is characterised by the formation T[middot]AT and C
+
[middot]GC triplets (
5
-
7
), while the antiparallel motif consists of G[middot]GC, A[middot]AT and T[middot]AT triplets (
8
-
11
). Because of the requirement for protonation of third strand cytosines, the
parallel motif requires conditions of low pH, while antiparallel structures are stable at physiological pHs. Both motifs are stabilised by divalent metal ions, particularly magnesium (
12
,
13
).
Although third strand oligonucleotides possess exquisite sequence recognition
properties, their binding is not strong compared with that of DNA duplexes. One
means of increasing their binding strength is to use triplex-specific ligands, which bind to triplex but not duplex DNA. Several such
compounds have now been described including a series of benzopyridoindole
derivatives (
14
-
19
), coralyne (
20
,
21
), 2,6-disubstituted amidoanthraquinones (
22
) and a series of unfused naphthylquinolines (
23
-
25
). Most studies on these compounds have examined their interaction with pyrimidine-rich (parallel) third strands, though it has recently been shown that BePI can stabilise antiparallel triplexes with GT- but not GA-containing oligonucleotides (
19
).
In previous studies we have shown that the naphthylquinoline compound (Fig.
1
a) reduces the concentration of third strand oligonucleotides required to
generate DNase I footprints by up to 200-fold (
24
,
25
). In these experiments we examined the interaction of T
5
C
5
and C
5
T
5
with the target sites A
6
G
6
[middot]C
6
T
6
and G
6
A
6
[middot]T
6
C
6
respectively (
24
). In the absence of the triplex-binding ligand these oligonucleotides do not generate DNase I footprints, since
they are short and produce blocks of five contiguous C
+
[middot]GC residues. In the presence of 10 [mu]M ligand, footprints are observed with oligonucleotide
concentrations as low as 0.2 [mu]M. These studies suggested that the naphthylquinoline compound had a greater
effect on parallel than antiparallel triplexes (
24
). In this paper we examine the sequence selectivity of this ligand, comparing
its effect on C
+
[middot]GC and T[middot]AT triplets, and extend our studies on its effect on the
formation of antiparallel triplets using either GA- or GT-containing oligonucleotides. These studies have used the same target sites as in previous studies (
24
,
26
,
27
), attempting to generate the complexes shown in Figure
1
b and c.
Oligonucleotides were purchased from Oswel, stored at -20oC and diluted to working concentrations in appropriate buffers, as
indicated, immediately before use. The acridine-linked oligonucleotides Acr-T
5
C
5
and Acr-C
5
T
5
, in which the (2-methoxy, 6-chloro, 9-amino) acridine is linked to the 5'-end by a pentamethylene chain were gifts from Dr M. J. McLean, Cambridge
Research Biochemicals. The naphthylquinoline triplex binding ligand (Fig.
1
a) was prepared as previously described (
23
) and was stored at -20oC as a stock solution in dimethylsulphoxide, and diluted into
aqueous buffers immediately before use. The pK of this compound was determined
to be 7.1 by measuring the change in absorbance at 330 nm as a function of pH,
measuring the pH on an Accumet 910 pH-meter using an Accumet microprobe glass electrode with an Ag/AgCl reference. The buffers used in the footprinting experiments were either 50 mM sodium acetate pH 5.5 containing 5 mM MgCl
2
, or 10 mM Tris-HCl pH 7.5 containing 5 mM MgCl
2
.
The preparation of plasmids pAG1 and pGA1 has been previously described (
26
,
27
). These contain the sequences A
6
G
6
[middot]C
6
T
6
and G
6
A
6
[middot]T
6
C
6
respectively, cloned into the
Bam
HI site of plasmid pUC18. Radiolabelled DNA fragments were prepared by digesting the fragments with
Hin
dIII, labelling at the 3'-end of the
Hin
dIII site with [[alpha]-
32
P]dATP using reverse transcriptase, and cutting again with
Eco
RI. The labelled fragments were separated from the remainder of the plasmid on 8% non-denaturing polyacrylamide gels. The inserts are oriented so that this procedure visualises the
purine strand of pAG1 and the pyrimidine strand of pGA1.
Radiolabelled DNA fragments (1.5 [mu]l) containing the target sites were mixed with 1.5 [mu]l of the oligonucleotide and 1.5 [mu]l of triplex- binding ligand dissolved in buffer as indicated. The complexes
were incubated at room temperature for at least 30 min, sufficient to achieve equilibrium, before adding 2 [mu]l DNase I diluted in 1 mM MgCl
2
, 1 mM MnCl
2
, 20 mM NaCl. Longer incubation times did not alter the digestion patterns. The enzyme concentration was chosen so as to
ensure single hit kinetics and was typically 0.02 U/ml at pH 7.5 and 0.2 U/ml at pH 5.0. The digestion was stopped after 1 min by the
addition of 4 [mu]l DNase I stop solution (80% formamide containing 10 mM EDTA). Samples were
heated at 100oC for 3 min before electrophoresis.
Products of the digestion were separated on 10% polyacrylamide gels, containing 8 M urea and run at 1500 V for ~2 h. The gels were then fixed in 10% (v/v) acetic acid before drying at 80oC and subjecting to autoradiography at -70oC using an intensifying screen. Bands were assigned by
comparison with Maxam-Gilbert sequencing lanes specific for purines.
We have previously shown that the naphthylquinoline triplex-binding ligand (Fig.
1
a) promotes the binding of T
5
C
5
to the target site A
6
G
6
[middot]C
6
T
6
(
24
). This oligonucleotide alone does not produce DNase I footprints even at concentrations as high as 50 [mu]M, but generates clear footprints in the presence of 10 [mu]M ligand which persist to oligonucleotide concentrations as low as 0.2 [mu]M. We have investigated whether this stabilization arises from
selective interaction of the ligand with either T[middot]AT or C
+
[middot]GC triplets by examining its effect on the binding of T
6
C
2
and T
2
C
6
to the same target site. The results of these experiments are shown in Figure
2
. T
6
C
2
(left hand panel) should form a complex containing 6* T[middot]AT and 2* C
+
[middot]GC triplets while T
2
C
6
(right hand panel) should form 2* T[middot]AT and 6* C
+
[middot]GC triplets. It can be seen that neither oligonucleotide affects the
DNase I cleavage pattern in the absence of added ligand, even at a
concentration of 50 [mu]M. In the presence of 10 [mu]M ligand T
6
C
2
produces a clear footprint, covering the upper part of the A
6
G
6
target site, which persists down to a concentration of 1 [mu]M oligonucleotide. A region of enhanced cleavage is evident at the lower (3') end of the footprint at the triplex-duplex junction. This effect has previously been observed with T
5
C
5
(
24
,
26
), which generates a similar enhancement at its triplex-duplex junction, located three bands further down the gel. The position
of this enhancement, at the boundary of the predicted triplex-duplex junction provides evidence that the terminal C
+
[middot]GC triplets actively contribute to the interaction, and that the
footprint is not merely caused by a strong interaction of the T[middot]AT triplets. This is further evidenced by the observation that this
ligand-induced footprint is pH dependent and is not apparent above pH 5.5 (not
shown). In contrast to these results no footprint is observed with T
2
C
6
, even at a concentration of 50 [mu]M in the presence of 10 [mu]M triplex-binding ligand (right hand panel). Higher concentrations of the
triplex-binding ligand (up to 300 [mu]M) did not facilitate the binding of T
2
C
6
. Since the ligand promotes the formation of a complex containing 6* T[middot]AT and 2* C
+
[middot]GC triplets but not 2* T[middot]AT and 6* C
+
[middot]GC triplets, this strongly suggests that it selectively stabilises the T[middot]AT triplet. No footprints are produced with T
6
, lacking the two terminal cytosines (not shown), presumably because this short triplex is not stable
under these conditions.
The sequences A
6
G
6
[middot]C
6
T
6
and G
6
A
6
[middot]T
6
C
6
can in theory be targeted with G
5
T
5
and T
5
G
5
respectively, forming antiparallel triplexes containing 5* G[middot]GC and 5* T[middot]AT triplets, which should be stable at
physiological pH, generating the structures shown in Figure
1
b and c. However, these oligonucleotides do not generate DNase I footprints,
presumably because they are too short to form stable complexes. In addition,
previous studies with derivatives possessing 5'-acridine moieties have shown that although Acr-G
5
T
5
binds to A
6
G
6
[middot]C
6
T
6
no footprints are observed with Acr-T
5
G
5
and G
6
A
6
[middot]T
6
C
6
(
27
). We have previously shown that the complexes with unmodified oligonucleotides can not be stabilised by 10 [mu]M of the naphthylquinoline triplex-binding ligand (
24
). Figure
3
shows the effect of higher ligand concentrations on the formation of these
complexes. Looking first at the results with the target site A
6
G
6
[middot]C
6
T
6
(left hand panel) it can be seen that the ligand induces a footprint with 50 [mu]M oligonucleotide, which is centred around the target site and which
persists down to a ligand concentration of 20 [mu]M. Similar experiments (not shown) maintaining the ligand concentration at
20 [mu]M and varying the concentration of G
5
T
5
showed that the footprint was stable down to ~1 [mu]M oligonucleotide. The right hand panel of Figure
3
shows the effect of the ligand on the interaction between 50 [mu]M T
5
G
5
and G
6
A
6
[middot]T
6
C
6
. In this case the ligand only induces a triplex footprint at concentrations of
75 [mu]M and above, at which concentration the ligand itself affects the DNase I
cleavage pattern in regions outside the triplex binding site. Our previous
studies suggested that, although the ligand alone does not affect DNase I
cleavage patterns at concentrations of 20 [mu]M and below, it binds to duplex DNA at higher concentration at which it is
no longer selective for triplex structures (
24
). Nonetheless we find that on maintaining the ligand concentration at 75 [mu]M, the footprint persists down to an oligonucleotide concentration of ~1 [mu]M (not shown). This is the first instance in which we have been
able to demonstrate successful triple helix formation at this site with GT-containing oligonucleotides. It therefore appears that the ligand can stabilise the formation of antiparallel triplexes with short
unmodified GT-containing oligonucleotides, though it binds to these complexes less well than the corresponding parallel structures formed with pyrimidine containing oligonucleotides.
Antiparallel triplexes can also be formed with GA-containing oligonucleotides, generating G[middot]GC and A[middot]AT triplets, though we have previously failed to produce
complexes between G
5
A
5
and A
6
G
6
[middot]C
6
T
6
or between A
5
G
5
and G
6
A
6
[middot]T
6
C
6
(
24
). We find that the triplex-binding ligand does not promote the formation of either complex at pH 7.5.
At pH 5.5 there is also no interaction between G
5
A
5
and A
6
G
6
[middot]C
6
T
6
, though the ligand promotes the formation of a complex between A
5
G
5
and G
6
A
6
[middot]T
6
C
6
(Fig.
5
). In this figure it can be seen that 20 [mu]M ligand promotes the formation of a footprint which is stable down to 10 [mu]M A
5
G
5
.
Most of the triplex binding ligands which have been described are thought to act
by an intercalative mechanism. We therefore tested whether the ability to
stabilise these intermolecular triplexes is a general property of intercalators
by examining the effect of ethidium on their formation. There have been several
reports that ethidium can bind to DNA triple helices (
30
-
33
), though it is not clear whether it can actually promote their formation. We
find that 10 [mu]M ethidium does not promote the formation of any of the intermolecular
triplexes shown in Figure
1
b and c confirming that the ability to stabilise these triplexes is not a
general property of all intercalating agents. Since these short unmodified
triplexes are not stable in the absence of added ligand we can not determine
whether ethidium merely fails to promote complex formation or if it actively
destabilizes the complexes. We have attempted to address this question by
examining the effect of ethidium on the interaction of short acridine-linked oligonucleotides to these two target sites. The results of this experiment are presented in Figure
6
. It can be seen that, as previously reported (
24
,
26
), both acridine-linked oligonucleotides produce clear footprints at their target sites. These footprints are abolished by ethidium concentrations of 20 [mu]M and above, producing DNase I cleavage patterns which are similar to
those seen with ethidium alone.
The results presented in this paper suggest that, within the parallel motif, the triplex-binding ligand selectively stabilises the formation of T[middot]AT rather than C
+
[middot]GC triplets. It seems unlikely that this selectivity arises from the
formation of specific hydrogen bonds. Since T[middot]AT and C
+
[middot]GC triplets are isohelical the area of overlap with the two types of
base triplets should be similar and stacking interactions are unlikely to
account for this difference. We suspect that the sequence selectivity arises
from charge repulsion between the protonated cytosine and the cationic triplex-binding ligand. In this regard it is worth noting that all the
intercalative triplex binding ligands described to date possess cationic
chromophores, and so may display the same sequence selectivity. The only
exception to this is coralyne which has been reported to bind to C
+
[middot]GC triplets as well as T[middot]AT (
20
,
21
). It is worth emphasising that, although triplex-binding ligands have been shown to stabilise complexes containing triplet mismatches (
25
), the cytosines in the interaction with T
6
C
2
are actively contributing to the integrity of the complex, even though they may
not be stabilised by the ligand. This is evident from the pH dependency of the
complex and from the size of the footprint, which terminates exactly three
bases earlier than the corresponding footprint with T
5
C
5
. It has previously been shown that the naphthylquinoline triplex-binding ligand can promote the formation of alternating G[middot]TA and T[middot]AT triplets when these are tethered to an adjacent block
of T[middot]AT triplets (
34
). The results with T
6
C
2
show that a block of six consecutive T[middot]AT triplets is unable to stabilise the interaction of two adjacent C[middot]GC triplets at pHs above 5.5.
These results demonstrate that the triplex-binding ligand is more potent at stabilising the formation of antiparallel
triplexes with GT- than GA-containing oligonucleotides. A similar effect has recently been
noted for BePI (
19
). This difference suggests that the ligand is interacting with T[middot]AT rather than G[middot]GC. The apparent preference for T[middot]AT over G[middot]GC or A[middot]AT is probably due to differences in the
conformations adopted by the various triplets which are not isomorphous (
28
). Moreover, since the antiparallel R[middot]RY triplets have a larger surface area than T[middot]AT they may not properly accommodate the ligand and generate
poor overlap between the ligand and the bulkier third strand purine residues. We suggest that stabilisation of A[middot]AT and G[middot]GC triplets may require ligands with larger chromophores. An alternative explanation for
the different effect on GT- and GA-containing oligonucleotides is that under these conditions T[middot]AT is more stable than A[middot]AT and is therefore more readily affected by the
ligand.
Although the triplex-binding ligand has a large effect on the formation of antiparallel triplexes with GT-containing oligonucleotides at the target site A
6
G
6
[middot]C
6
T
6
, it has a much smaller effect on similar complexes formed at the target site G
6
A
6
[middot]T
6
C
6
in which the order of G[middot]GC and T[middot]AT triplets has been reversed. This is similar to the relative
affinities of acridine-linked GT-containing oligonucleotides at these targets sites (
27
) for which no footprints were observed with Acr-T
5
G
5
and G
6
A
6
[middot]T
6
C
6
. This suggests that the order of G[middot]GC and T[middot]AT triplets is an important factor in the formation of these
complexes. It is possible that these results may be affected by the formation
of stable secondary structures in the GT-containing oligonucleotides. However this can not be the sole explanation
since Acr-T
5
G
5
is able to form a stable alternate strand triplex (
35
). A possible explanation for this difference concerns the different
conformations adopted by each triplet. T[middot]AT and G[middot]GC triplets are not isomorphous (
28
) and there will be a discontinuity in the conformation of the third strand at
the ApG or GpA junction. An alternative explanation is based on the suggestion
that GT-containing oligonucleotides anneal to their target sites in an asymmetric fashion, proceeding in a 3'-5' direction (
36
). This asymmetric nucleation may explain our results if we assume that the ligand has a greater effect on antiparallel T[middot]AT than G[middot]GC triplets. The initial interaction of G
5
T
5
with A
6
G
6
[middot]C
6
T
6
would be stabilised by ligand binding to the 5* T[middot]AT triplets at the 3'-end of the oligonucleotide. In contrast, with T
5
G
5
and G
6
A
6
[middot]T
6
C
6
the initial annealing complex will only contain G[middot]GC triplets and so will be affected by the ligand to a lesser extent.
It should also be noted that if a triplex is inherently unstable, then adding a
ligand may not be sufficient to induce a footprint, even if it binds with the
same affinity as to other more stable triple helices. Since the stabilisation
will not be a linear function of ligand concentration, as it is essentially a
trimolecular interaction, the concentration of ligand required to generate a
clear footprint at different oligonucleotide concentrations may not provide a rigorously quantitative estimate of its affinity for different triplets.
It is interesting to note that at low pH the binding of A
5
G
5
to its target site is facilitated by the ligand whereas no interaction with G
5
A
5
was observed at any ligand concentration. Once again the order of the blocks of
triplets appears to influence the stability of the complexes. As discussed
above for the GT-containing oligonucleotides this could arise from differences between GpA
and ApG junctions, or from an asymmetric annealing process.
We have previously shown that parallel triplexes containing T[middot]AT and C
+
[middot]GC triplets are stabilised by the ligand regardless of the order of the
triplets (
24
). In contrast the effect of the ligand on T[middot]AT and G[middot]GC triplets depends on the order of blocks of G[middot]GC and T[middot]AT triplets. Two explanations can be offered for
this difference. Firstly parallel C
+
[middot]GC and T[middot]AT triplets are isomorphous, whereas antiparallel G[middot]GC, A[middot]AT and T[middot]AT are not. Secondly the initial
nucleation reaction can occur in either direction for the parallel motif (
37
), but proceeds in the 3'-5' direction for the antiparallel structures (
36
). It is also interesting to note that parallel triplexes induce enhanced DNase
I cleavage at the 3'-end of the purine strand at the triplex-duplex junction, while no such changes are seen with the
antiparallel motif. The conformation of the underlying duplex therefore appears
to be different in the two motifs. Although the ligand has different effects on
parallel and antiparallel complexes generated at the target site G
6
A
6
[middot]T
6
C
6
we find that, at low pHs, it is equipotent at stabilising antiparallel (GT-containing) and parallel (CT-containing) complexes at the target site A
6
G
6
[middot]T
6
C
6
. This suggests that it has a similar effect on parallel and antiparallel T[middot]AT triplets.
The improvement in the ability of the ligand to stabilise antiparallel triplexes at low pH was unexpected, since the formation of these complexes
should be pH independent. This presumably reflects changes in the charge on the
ligand. The terminal nitrogen on the aminoalkyl side chain should be fully
charged at all pHs, whereas the ring nitrogen has a pK of 7.1. Changing the pH
from 7.0 to 5.5 will therefore double the concentration of the protonated
species. These results suggest that the charge on the ring nitrogen is
important for efficient interaction with the triplets. However, a positively charged intercalating moiety is not a sufficient requirement for triplex-stabilisation since ethidium does not promote the formation of complexes
with unmodified oligonucleotides and destabilises the interaction with acridine-linked compounds. The requirement for a positive charge on the
intercalating ring may imply that this series of compounds will always be
selective for the T[middot]AT triplet as a result of charge repulsion with C
+
[middot]GC.
This work was supported by grants from the Cancer Research Campaign and the
Medical Research Council.
*To whom correspondence should be addressed. Tel: +44 1703 594374; Fax: +44 1703
594319; Email: KRF1@soton.ac.uk
REFERENCES
Return