DNA sequence specificity of triplex-binding ligands

Authors


K. R. Fox, Division of Biochemistry & Molecular Biology, School of Biological Sciences, University of Southampton, Bassett Crescent East, Southampton SO16 7PX, UK.
Fax: + 44 23 80594459, Tel.: + 44 23 80594374,
E-mail: K.R.Fox@soton.ac.uk

Abstract

We have examined the ability of naphthylquinoline, a 2,7-disubstituted anthraquinone and BePI, a benzo[e]pyridoindole derivative, to stabilize parallel DNA triplexes of different base composition. Fluorescence melting studies, with both inter- and intramolecular triplexes, show that all three ligands stabilize triplexes that contain blocks of TAT triplets. Naphthylquinoline has no effect on triplexes formed with third strands composed of (TC)n or (CCT)n, but stabilizes triplexes that contain (TTC)n. In contrast, BePI slightly destabilizes the triplexes that are formed at (TC)n (CCT)n and (TTC)n. 2,7-Anthraquinone stabilizes (TC)n (CCT)n and (TTC)n, although it has the greatest effect on the latter. DNase I footprinting studies confirm that triplexes formed with (CCT)n are stabilized by the 2,7-disubstituted amidoanthraquinone but not by naphthylquinoline. Both ligands stabilize the triplex formed with (CCTT)n and neither affects the complex with (CT)n. We suggest that BePI and naphthylquinoline can only bind between adjacent TAT triplets, while the anthraquinone has a broader sequence of selectivity. These differences may be attributed to the presence (naphthylquinoline and BePI) or absence (anthraquinone) of a positive charge on the aromatic portion of the ligand, which prevents intercalation adjacent to C+GC triplets. The most stable structures are formed when the stacked rings (bases or ligand) alternate between charged and uncharged species. Triplexes containing alternating C+GC and TAT triplets are not stabilized by ligands as they would interrupt the alternating pattern of charged and uncharged residues.

Abbreviation
BePI

3-methoxy-7H-8-methyl-11[(3′-amino)propyl- amino]-benzo[e]pyrido[4,3-b]indole

The formation of intermolecular DNA triple helices offers a means for designing agents which can bind to specific DNA sequences [1–6]. In this approach, a third strand oligonucleotide binds in the major groove of duplex DNA, forming specific hydrogen bond contacts with substituents on the purines of the duplex base pairs. In these structures the third strand can run either parallel or antiparallel to the duplex purine strand. Parallel triplexes, which have been most widely studied, consist of TAT and C+GC triplets, while the antiparallel motif contains GGC and AAT or TAT triplets. Although triplex-forming oligonucleotides bind to their duplex targets with considerable sequence selectivity, their binding may not be strong. Several strategies have therefore been used to improve their affinity, including the use of base and backbone analogues [7–10], and tethering DNA binding agents such as acridine [11,12] or psoralen [13] to the end of the oligonucleotide. An alternative strategy uses ligands which bind selectively to triplex (not duplex) DNA and which therefore perturb the equilibrium in the direction of triplex formation. Several such ligands have been described (reviewed in [14]) including 3-methoxy-7H-8-methyl-11[(3′-amino)propylamino]-benzo[e]pyrido[4,3-b]indole (BePI) and its derivatives [15–18], coralyne [19–21], naphthylquinoline derivatives [22–26] and bis-substituted amidoanthraquinones [27,28]. Although most studies with these compounds have examined their interaction with parallel triplets, they also stabilize the antiparallel motif [24,25,29].

Although many ligands have been used to stabilize triple helical DNA, little is known about their sequence selectivity, in particular whether they have similar affinities for TAT and C+GC triplets. Coralyne was originally thought to possess little sequence selectivity [19], but was later shown to be selective for TAT triplets [21]. BePI favours adjacent TAT triplets within a parallel triplex [15], but can also stabilize antiparallel triplexes consisting of both TAT and GGC triplets [29]. Studies with a naphthylquinoline compound (Fig. 1) indicated that at low pH this ligand stabilizes TAT triplets in preference to C+GC triplets [25]. This preference is thought to be due to the cationic charge on the naphthylquinoline group, which hinders intercalation adjacent to a protonated cytosine. Other studies have shown that isolated C+GC triplets impart a greater stability than TAT triplets [30–32], although blocks of contiguous C+GC triplets are destabilizing [33]. It is therefore possible that the apparent selectivity of most ligands for TAT triplets is a consequence of the lower stability of this triplet, making it easier to stabilize further.

Figure 1.

Chemical structures of the 2,7-disubstituted anthraquinone, naphthylquinoline and sequences of the 110-base pair fragment from tyrT(35–59) and the 25-base pair oligopurine tract together with the four oligonucleotides used. (A) Chemical structures of the 2,7-disubstituted anthraquinone. (B) Chemical structure of the naphthylquinoline. (C) Sequence of the 110-base pair fragment from tyrT(35–59). The 25-base pair oligopurine tract is underlined. In all these studies the fragment was labelled at the 3′-end of the lower strand. (D) Sequence of the 25 base pair oligopurine tract (boxed) together with the four oligonucleotides which are targeted to different regions.

In this paper we use DNase I footprinting and fluorescence melting experiments to assess the effects of ligands on triplexes of different base composition. For the footprinting experiments we have prepared a novel fragment that can be targeted with oligonucleotides of different base composition. TyrT(35–59) contains a 25-base pair oligopurine tract, which can be targeted with three different oligonucleotides as shown in Fig. 1. Oligo 1 (CCT)n generates a triplex containing pairs of adjacent C+GC triplets which are separated by single T≅AT triplets, while oligo 2 (CCTT)n produces pairs of adjacent TAT and C+GC. Oligo 3 (CT)n generates a triplex containing alternating TAT and C+GC triplets. We would expect that AT-selective ligands should only potentiate the binding of oligo 2, as this is the only triplex containing adjacent TAT triplets. We have compared the effects of two triplex-binding ligands, a naphthylquinoline and a 2,7-disubstituted amidoanthraquinone, on each of these triplexes. These experiments were augmented with fluorescence melting studies with inter- and intramolecular triplexes with different sequence arrangements. The sequences of these fluorescently labelled oligonucleotides are shown below (Results, Fluorescence melting studies).

Materials and methods

Chemicals and enzymes

Oligonucleotides were purchased from Oswel DNA Service (Southampton, UK). These were stored in water at −20 °C and diluted to working conditions immediately before use. The sequences of the intermolecular triplexes generated in the footprinting experiments are shown in Fig. 1D, while the synthetic oligonucleotides that were used for the fluorescence melting studies are shown below (Results, Fluorescence melting studies). To avoid any potential problems with misannealing of the fluorescently labelled intermolecular triplexes, we used intramolecular duplexes in which the two strands were connected by a single hexaethylene glycol moiety (H). The fluorophore (fluorescein) was incorporated at the 5′-end of the duplex DNA, and the quencher (Methyl red) was attached at the 5′-end of the third strand oligonucleotide. The 15-mer sequences were chosen so as to generate triplexes with different arrangements of C+GC and TAT triplets and are based around repeats of (CCT)n (CT)n (CTT)n, and Tn. The symmetrical repeating structure of each complex was deliberately broken so as to prevent strand slippage. For the fluorescently labelled intramolecular triplexes the three strands were also linked with hexaethylene glycol moieties. These sequences were chosen to produce 14-mer intramolecular duplexes with an internal fluorescein label within the purine-rich strand. The quencher was attached to the 5′-end of the short third strand which generates a six-base triplex, similar to that used in our previous studies [34]. The third strand was shorter than the underlying duplex, to ensure that the duplex–single-strand and triplex–duplex transitions occur at different temperatures.

The 2,7-disubstituted amidoanthraquinone with a pyrrolidine end group was prepared as the water-soluble addition salt as using standard procedures as described previously [35,36]. The naphthylquinoline triplex-binding ligand was a gift from L. Strekowski (Department Chemistry, Georgia State University, Atlanta, USA). This was stored at a stock concentration of 20 mm in dimethylsulfoxide. BePI was from Sigma.

DNA fragment for footprinting studies

For the footprinting studies, we prepared a novel derivative of the tyrT DNA fragment which contains a 25-base oligopurine tract. This fragment [designated tyrT(35–59)] was derived from tyrT(43–59) by PCR site-directed mutagenesis, and was designed so that different parts of the tract contain (CCT)n (CCTT)n and (TC)n repeats. The sequence of this fragment is shown in Fig. 1, together with the four 9-mer oligonucleotides which form specific triplexes with different regions of this target. The radiolabelled DNA fragment was prepared by digesting the plasmid with EcoRI and AvaI and was labelled at the 3′-end of the EcoRI site using reverse transcriptase and [32P]dATP[αP]. The labelled 110-base pair DNA fragment was separated from the remainder of the plasmid DNA on an 8% (w/v) nondenaturing polyacrylamide gel. The isolated DNA was dissolved in 10 mm Tris/HCl pH 7.5 containing 0.1 mm EDTA to give about 10–20 c.p.s.·µL−1 as determined on a hand held Geiger counter (< 10 nm). For quantitative footprinting experiments, the absolute DNA concentration is not important so long as it is lower than the dissociation constant of the DNA binding compound.

DNase I footprinting

Radiolabelled DNA (1.5 µL) was mixed with 1.5 µL oligonucleotide and 1.5 µL triplex binding ligand. The ligand and oligonucleotide were both dissolved in 50 mm sodium acetate (pH 5.0) containing 10 mm MgCl2. The concentrations refer to conditions in the final reaction mixture. These mixtures were equilibrated at 20 °C for at least 2 h. The samples were digested by adding 2 µL DNase I (typically 0.01 U·mL−1) dissolved in 20 mm NaCl containing 2 mm MgCl2 and 2 mm MnCl2. The reaction was stopped after 1 min by adding 5 µL 80% formamide containing 10 mm EDTA, 10 mm NaOH and 0.1% (w/v) Bromophenol blue.

Gel electrophoresis

The products of digestion were separated on 9% polyacrylamide gels containing 8 m urea. Samples were heated to 100 °C for 3 min, before rapidly cooling on ice and loading onto the gel. Polyacrylamide gels (40 cm long, 0.3 mm thick) were run at 1500 V for about 2 h and then fixed in 10% (v/v) acetic acid. These were transferred to Whatman 3MM paper and dried under vacuum at 80 °C. The dried gels were either exposed to X-ray film at −70 °C using an intensifying screen, or were subjected to phosphorimaging using a Molecular Dynamics STORM phosphorimager.

Quantitative analysis of footprinting data

The intensity of bands within each footprint was estimated using imagequant software. These were normalized by comparison with a region for which DNase I cleavage was not affected. Footprinting plots [37] were constructed from these data and C50 values, indicating the oligonucleotide concentration which reduces the band intensity by 50%, were calculated by fitting a simple binding curve to the data.

Fluorescence melting studies

Fluorescence melting profiles were determined by using a Roche LightCycler as described previously [34]. The principle of these experiments is that when a triplex is formed the fluorophore and quencher are in close proximity and the fluorescence is quenched. On denaturing the complex, the fluorophore and quencher are separated and there is a large increase in fluorescence. We have previously used this method for examining the properties of DNA triplexes [34,38] and quadruplexes [39]. This method is especially useful for studying triplex formation as only the triplex–duplex transition is observed, and the analysis is not complicated by the melting of the underlying duplex. Samples were prepared in 50 mm sodium acetate pH 5.0 containing 150 mm NaCl. For the intermolecular triplexes each sample (20 µL) contained 0.25 µm duplex DNA and 4 µm triplex forming oligonucleotide, while for the intramolecular triplexes the strand concentration was 0.25 µm. The complexes were denatured by heating to 95 °C at a rate of 0.1 °C·s−1 and maintained at this temperature for 5 min before cooling to 30 °C at 0.1 °C·s−1. Samples were then held at 30 °C for 5 min before melting again by heating to 95 °C at 0.1 °C·s−1. The fluorescence was recorded during both melting and annealing phases. The LightCycler excites the samples at 488 nm, and the emission was measured at 520 nm. Tm values were determined from the first derivatives of the melting profiles using the Roche lightcycler software and were reproducible to within 0.5 °C. Unless otherwise stated the Tm values quoted refer to the second melting transition. The properties of the intermolecular triplexes have been described previously [38].

Results

DNase I footprinting

Oligo 1. Fig. 2 shows the interaction of oligo 1(TCCTCCTCC) with its target site on tyrT(35–59) in the presence and absence of the two triplex-binding ligands. These experiments were performed at pH 5.0 to ensure protonation of the third strand cytosines. As this triplex does not contain any adjacent TAT triplets it should not be stabilized by AT-selective triplex-binding ligands. The oligonucleotide alone (left panel) produces a clear footprint at its target site, which extends to an oligonucleotide concentration of about 5 µm. This footprint extends a few bases beyond the 5′-(upper) end of the target site and is accompanied by enhanced cleavage at the 3′-(lower) end, coincident with the triplex–duplex junction. Quantitative analysis of the concentration dependence of the footprint yielded a C50 value of 2.8 ± 0.6 µm.

Figure 2.

DNase I footprints showing the interaction of oligo 1(5′-TCCTCCTCC) with its target site on tyrT(35–59). The first panel (Left) shows the pattern produced by the ligand alone; the second and third panels show the pattern in the presence of 3 and 1 µm naphthylquinoline, while the last two panels show the pattern in the presence of 3 and 1 µm 2,7-anthraquinone. The bracket shows the position of the 25-base oligopurine tract, while the filled boxes show the location of the 9-base target site for this oligonucleotide. The track labelled GA is a Maxam–Gilbert marker specific for purines. Oligonucleotide concentrations (µm) are shown at the top of each gel lane.

Similar experiments were performed in the presence of 1, 3 and 10 µm of the naphthylquinoline and the 2,7-disubstituted anthraquinone triplex-binding ligands. The second and third panels of Fig. 2 show the footprints obtained in the presence of 3 and 1 µm of the naphthylquinoline triplex-binding ligand, respectively. In each case the oligonucleotide produces a clear footprint which is evident only at concentrations above 2 µm. Quantitative analysis of these patterns yields the footprinting plots shown in Fig. 3A, generating the C50 values presented in Table 1. It can be seen that these values are very similar to that for the oligonucleotide alone, suggesting that the ligand does not significantly stabilize this triplex (even at a concentration of 10 µm).

Figure 3.

Footprinting plots showing the concentration dependence of the footprints obtained with the three oligonucleotides at their target sites in tyrT(35–59), in the presence and absence of triplex-binding ligands. (A) Oligo 1 plus naphthylquinoline: •, oligonucleotide alone with no added ligand; ○, 3 µm ligand; ▿, 1 µm ligand. (B) Oligo 1 plus 2,7 anthraquinone: ▵, 10 µm ligand; ○, 3 µm ligand; ▿, 1 µm ligand. (C) Oligo 2 plus naphthylquinoline: ▵, 10 µm ligand; ○, 3 µm ligand; ▿, 1 µm ligand. (D) Oligo 2 plus 2,7-anthraquinone: •, oligonucleotide alone with no added ligand; ▵, 10 µm ligand; ○, 3 µm ligand; ▿, 1 µm ligand. (E) Oligo 3: •, oligonucleotide alone with no added ligand; ▵, 3 µm anthraquinone; ○, 3 µm naphthylquinoline. (F) Oligo 4: • oligonucleotide alone with no added ligand; ▵, 3 µm anthraquinone, ○, 3 µm naphthylquinoline.

Table 1. C50 and KL values for the interaction of the various triplex forming oligonucleotides with their target sites on tyrT (35–59). C50 values (µm), corresponding to the oligonucleotide concentration which reduces the intensity of bands within the target site by 50%, were determined as described in Methods.
OligoLigandC50m)
Oligo 1No ligand2.8 ± 0.6
10 µm AQ0.09 ± 0.02
3 µm AQ0.25 ± 0.07
1 µm AQ0.45 ± 0.09
10 µm NQ2.1 ± 0.5
3 µm NQ1.5 ± 0.3
1 µm NQ2.0 ± 0.4
Oligo 2No ligand1.0 ± 0.2
10 µm AQ0.06 ± 0.02
3 µm AQ0.12 ± 0.05
1 µm AQ0.38 ± 0.08
10 µm NQ0.12 ± 0.03
3 µm NQ0.43 ± 0.14
1 µm NQ0.73 ± 0.23
Oligo 3No ligand0.14 ± 0.04
3 µm AQ0.10 ± 0.04
1 µm AQ0.14 ± 0.07
3 µm NQ0.12 ± 0.04
1 µm NQ0.12 ± 0.04
Oligo 4No ligand0.9 ± 0.2
3 µm AQ0.6 ± 0.2
3 µm NQ0.7 ± 0.2

The final two panels of Fig. 2 show the footprinting patterns obtained in the presence of 3 and 1 µm of the 2,7-anthraquinone. Once again clear footprints are evident at the target site, but require lower oligonucleotide concentrations than in the absence of ligand. In the presence of 3 and 1 µm of the anthraquinone the footprints extend to oligonucleotide concentrations of 0.4 and 1 µm, respectively. Footprinting plots derived from these data are shown in Fig. 3B yielding the C50 values which are presented in Table 1. From these data it appears that the 2,7-disubstituted anthraquinone stabilizes this triplex in a concentration-dependent manner, whereas the naphthylquinoline does not affect the interaction with the oligonucleotide.

Oligo 2. Fig. 4 shows the interaction of oligo 2 (CTTCCTTCC) with its target site on tyrT(35–59) in the presence and absence of the two triplex-binding ligands. This oligonucleotide should generate a triplex containing two blocks of adjacent TAT triplets, which might provide a binding site for AT-selective ligands. The first panel shows the interaction with the oligonucleotide alone and shows a footprint at the target site. Quantitative analysis of the concentration dependence of this footprint (Fig. 3D,•) yields a C50 value of 1.0 ± 0.2 µm. Similar experiments in the presence of the triplex-binding ligands are shown in the other panels of Fig. 4, while footprinting plots derived from these data are shown in Fig. 3C and D. The footprints in the presence of 3 and 1 µm anthraquinone persist to about 0.2 and 0.4 µm, respectively, yielding C50 values of 0.12 and 0.38 µm. Similarly the footprints in the presence of 3 and 1 µm naphthylquinoline persist to 0.6 and 1 µm, yielding C50 values of 0.43 and 0.73 µm. Even lower C50 are generated in the presence of 10 µm ligands (Table 1).

Figure 4.

DNase I footprints showing the interaction of oligo 2(5′-CTTCCTTCC) with its target site on tyrT(35–59). The first panel (Left) shows the pattern produced by the ligand alone; the second and third panels show the pattern in the presence of 3 and 1 µm 2,7-anthraquinone, while the last two panels show the pattern in the presence of 3 and 1 µm naphthylquinoline. The bracket shows the position of the 25-base oligopurine tract, while the filled boxes show the location of the 9-base target site for this oligonucleotide. Tracks labelled GA are Maxam–Gilbert markers specific for purines. Oligonucleotide concentrations (µm) are shown at the top of each gel lane.

Oligos 3 and 4. Fig. 5 shows the interaction of oligo 4 (TCTCTCTCT) with its target site on tyrT(35–59) in the presence and absence of the two triplex-binding ligands. This oligonucleotide should generate a triplex which consists of alternating TAT and C+GC triplets. Robert and Crothers [40] predicted that such CT repeats should generate the most stable triplexes. It can be seen that this oligonucleotide generates a stable triplex at pH 5.0 in the absence of the ligands. This footprint, which is accompanied by enhanced cleavage at the 3′-(lower) end of the target site, persists to an oligonucleotide concentration between 0.2 and 0.1 µm. The footprinting plot for this interaction yields a C50 value of 0.14 ± 0.04 µm. Addition of either 3 µm of naphthylquinoline or anthraquinone has little effect on the apparent affinity for the third strand. Clear footprints are still visible which persist to very similar concentrations to those in the absence of ligand. The footprinting plots for this interaction are shown in Fig. 3E, yielding the C50 values shown in Table 1. From these values it can be seen that neither of these ligands significantly affects the triplex stability.

Figure 5.

DNase I footprints showing the interaction of oligo 3(5′-TCTCTCTCT) with its target site on tyrT(35–59). The first panel (Left) shows the pattern produced by the ligand alone (left hand lanes) and in the presence of 1 µm naphthylquinoline. The second panel shows the pattern in the presence of 3 µm naphthylquinoline, while the last panel shows the pattern in the presence of 3 µm 2,7-anthraquinone. The bracket shows the position of the 25-base oligopurine tract, while the filled boxes show the location of the 9-base target site for this oligonucleotide. The track labelled GA is a Maxam–Gilbert marker specific for purines. Oligonucleotide concentrations (µm) are shown at the top of each gel lane.

As oligo 3 binds tightly in the absence of ligand, we were concerned that the inherent stability of the complex might mask any effects due to the triplex-binding ligands. We therefore tested the interaction with a shorter 7-mer oligonucleotide (oligo 4), which lacks the two terminal thymine bases of oligo 3. We would expect oligo 4 to possess a lower triplex affinity than oligo 3 as a result of its shorter length. We avoided the use of even shorter oligos as these would be able to bind in more than one position. Footprinting experiments with this oligonucleotide (data not shown) confirmed that it bound less well than oligo 3, generating the footprinting plot shown in Fig. 3F (•), yielding a C50 value of 0.9 ± 0.2 µm. This is still a high affinity for such a short oligonucleotide and confirms the unusually high affinity of triplexes which consist of alternating TAT and C+GC triplets. Addition of either ligand failed to affect the concentration dependence of the footprints (Fig. 3F), yielding very similar C50 values (Table 1). It therefore appears that neither ligand binds to tracts of alternating TAT and C+GC triplets.

Fluorescence melting studies

We further examined the sequence specificity of these ligands by performing DNA melting experiments with fluorescently labelled synthetic oligonucleotides (Fig. 6). In these oligonucleotides the fluorophore and quencher are positioned so that they are in close proximity when a triplex is formed, and the fluorescence is quenched. When the triplex melts these groups are separated and there is large increase in fluorescence. Fluorescence melting curves showing the effects of naphthylquinoline, anthraquinone and BePI on the melting transitions of these triplexes are shown in Figs 7–9. Looking first at the results for the intramolecular triplexes (upper panels) it can be seen that, in the absence of added ligand, these all show unusual biphasic melting profiles in which the initial increase in fluorescence is followed by a decrease at higher temperatures. This second transition is less pronounced for the more stable triplexes (those containing C+GC triplets) and is not observed with the intermolecular triplexes (see below). We have suggested previously that this second transition arises from melting of the underlying duplex [34], since the time-averaged distance between the fluorophore and quencher is greater for the partially melted triplex than for the fully melted random coil. When the third strand dissociates the remaining duplex will be relatively rigid and hold the fluorophore and quencher apart. The Tm values for these transitions are presented in Table 2. It can be seen that the Tm of the second transition increases with increasing GC-content, as expected for melting of a DNA duplex, and is maximal for (CCT)2. In contrast the first melting transition is highest for (TC)3, consistent with the suggestion that regions of alternating C+GC and TAT triplets are unusually stable [40], and is similar to our previous work with intermolecular triplexes [38].

Figure 6.

Sequences of the fluorescently labelled inter- and intramolecular triplexes. Fluorescein (F) was incorporated as Fam-cap-dU, while the quencher methyl red (Q) was Methyl-Red-dR. H indicates the hexaethylene glycol joining the two duplex strands. In each case the Hoogsteen strand is shown in italics.

Figure 7.

Effect of the naphthylquinoline triplex-binding ligand on the fluorescence melting curves of the inter- and intramolecular triplexes. The upper panels show the results for the intramolecular triplexes, while the lower panels correspond to the intermolecular triplexes. The experiments were performed in 50 mm sodium acetate (pH 5.0) containing 150 mm sodium chloride, except for those labelled ‘+Mg’ where 50 mm MgCl2 was also included. In each case, the solid lines correspond to melting of the oligonucleotide alone, in the absence of added ligand, while the circles show the transitions in the presence of 10 µm ligand. For the triplexes that contain only TAT triplexes (first panel in each row), the curves shown correspond to ligand concentrations of 1, 2, 5 and 10 µm (intramolecular triplex T6) and 0.5, 1, 2, 5 and 10 µm (intermolecular triplex TTT), in each case the concentrations increase from left to right. The curves have been normalized to the same final fluorescence.

Figure 8.

Effect of the 2,7-disubstituted anthraquinone on the fluorescence melting curves of the inter- and intramolecular triplexes. The upper panels show the results for the intramolecular triplexes, while the lower panels correspond to the intermolecular triplexes. The experiments were performed in 50 mm sodium acetate (pH 5.0) containing 150 mm sodium chloride, except for those labelled ‘ +Mg’, where 50 mm MgCl2 was also included. In each case the solid lines correspond to melting of the oligonucleotide alone, in the absence of added ligand, the circles show the transitions in the presence of 10 µm ligand. For the triplexes that contain only TAT triplexes (first panel in each row) the curves shown correspond to ligand concentrations of 0.5, 2, 5 and 10 µm ligand; in each case the concentrations increase from left to right. The curves have been normalized to the same final fluorescence.

Figure 9.

Effect of BePI on the fluorescence melting curves of inter- and intramolecular triplexes. The upper panels show the results for the intramolecular triplexes, while the lower panels correspond to the intermolecular triplexes. The experiments were performed in 50 mm sodium acetate (pH 5.0) containing 150 mm sodium chloride, except for those labelled ‘+Mg’, where 50 mm MgCl2 was also included. In each case the solid lines correspond to melting of the oligonucleotide alone, in the absence of added ligand, while the circles show the transitions in the presence of 10 µm ligand. For the triplexes that contain only TAT triplexes (first two panels in each row) the curves shown correspond to ligand concentrations of 0.2, 0.5, 1, 2, 5 and 10 µm (intramolecular triplex T6), 2, 5 and 10 µm (intramolecular triplex T6 + Mg), 0.5, 1, 2, 5 and 10 µm (intermolecular triplex TTT) and 2, 5 and 10 µm (intermolecular triplex TTT + Mg). In each case the concentrations increase from left to right. The curves have been normalized to the same final fluorescence.

Table 2. Effects of triplex-binding ligand on the melting of intramolecular and intermolecular triplexes. The experiments were performed in 50 mm sodium acetate (pH 5.0) containing 150 mm NaCl. The concentration of the intramolecular triplexes was 0.25 µm. For the intermolecular triplexes hairpin duplex concentration was 0.25 µm, with 4 µm third strand. In each case the ligand concentration was 10 µm. For the intramolecular triplexes two melting transitions were observed: Tm1 corresponds to triplex→duplex, and Tm2 corresponds to duplex→single strands. ΔTm is equal to the Tm in the presence of ligand minus the Tm in its absence. Since the intermolecular TTT melts below 30 °C under these conditions the actual Tm values in the presence of ligands are shown. In the presence of BePI some melting transitions were unusually broad and are indicated by an asterisk ( *).
SequenceNo ligandNaphthylquinolineAnthraquinoneBePI
Tm1Tm2ΔTm1ΔTm2ΔTm1ΔTm2ΔTm1ΔTm2
Intra-
 T636.169.017.92.216.12.126.77.9
 T6 + Mg57.972.6 4.90.4 2.20.26.32.0
 (CTT)260.276.0 0.90 6.32.8−1.95.1
 (TC)367.279.6 0.60.6 3.11.4−7.7*4.9
 (CCT)264.781.2 1.31.3 3.61.2−5.8*5.5
 TmΔTmΔTmΔTm
Inter-
 TTT< 30Tm = 64.7Tm = 51.0Tm = 64.2
  TTT + Mg53.25.04.613.7
 CTT68.25.03.7 1.8
 CTCT75.20.91.3−4.2
 CCT74.6−0.40.4−5.3*

Figure 7 shows the effect of naphthylquinoline on these melting transitions and the ΔTms induced by 10 µm ligand are shown in Table 2. It can be seen that the ligand increases the melting temperature of triplexes that contain only TAT triplets (T6) and has little effect on the other triplexes. This effect is most pronounced in the absence of magnesium for which 10 µm ligand increases the melting temperature of the intramolecular complex by 17.9 °C. The intermolecular triplex that contains mainly TAT triplets (TTT) is unstable in the absence of ligand, and melts below 30 °C; addition of naphthylquinoline raises this Tm to 64.7 °C. The ligand has a much smaller effect on the more stable triplexes, and has no effect on triplexes that contain repeats of CCT and CTCT, though it stabilizes the intermolecular triplex CTT by the same amount as TTT in the presence of magnesium.

Figure 8 shows the results of similar experiments with the 2,7-anthraquinone. The overall pattern of these results is very similar to that seen with the naphthylquinoline and the greatest effect is again produced with the triplexes that contain TAT triplets. However, this ligand also produces a small, though significant, stabilization of CCT and CTCT triplexes.

For comparison we also examined the effect of BePI on melting of these fluorescently labelled triplexes, and the results are shown in Fig. 9 and Table 2. This ligand is one of the best characterized triplex-binding ligands and is well known to bind to triplexes that are rich in TAT triplets. It can be seen that BePI strongly stabilizes the TAT-containing triplets and is the most potent ligand at the intramolecular triplex T6. This activity is reduced on addition of magnesium, but is still greater than the other ligands under these conditions. BePI has a small stabilizing effect on the intermolecular triplex CTT. However, it destabilizes the intermolecular and intramolecular triplexes that contain repeats of CT or CCT, producing melting curves that are unusually shallow. These results suggest that BePI cannot bind adjacent to C+GC triplets and are consistent with previous observations that the effect of BePI decreases as the proportion of C+GC triplets increases.

Discussion

The results presented in this paper confirm that naphthylquinoline, anthraquinone and BePI, triplex-binding ligands, are much more effective at stabilizing triplexes that are rich in TAT triplets. This preference is most pronounced for BePI as the melting experiments showed a small destabilization of complexes containing a high proportion of C+GC triplets. Naphthylquinoline stabilizes only triplexes that contain adjacent TAT triplets and has no effect on the complexes that are formed with (CCT)n or (CT)n. In contrast, although the 2,7-disubstituted anthraquinone is most effective at stabilizing sequences that are rich in TAT triplets, it produces a small stabilization of all the triplexes examined, except those formed with oligos 3 and 4 in the footprinting experiments (generating triplexes that contain alternating TAT and C+GC triplets).

In general, all of these ligands are less effective at the more stable triplexes and ΔTm values with complexes TTT and T6 are much lower in the presence of magnesium, which selectively stabilizes the TAT triplet [38,41,42]. Nonetheless both naphthylquinoline and BePI still produce appreciable stabilization of this triplex in the presence of high concentrations of magnesium. It should be noted that these experiments were all performed at low pH in order to ensure cytosine protonation. We have previously reported the pH dependency of these triplexes [38] and have shown that, except for the complexes that contain only TAT triplets, they are not stable above pH 6.5. None of these ligands promotes triplex formation with C+GC-containing triplexes at pH 7.0. However, it will be interesting to discover whether they facilitate triplex formation at physiological pH with third strands containing cytosine analogues which have elevated pK values.

The quantitative DNase I footprinting experiments with oligo 1 (CCTCCTCCT), and the melting experiments with CCT and (CCT)2 clearly indicate that the naphthylquinoline ligand fails to stabilize this triplex. In contrast, the 2,7-disubstituted anthraquinone increases the melting temperatures, and lower the concentrations that are required to generate DNase I footprints. The inability of the naphthylquinoline ligand to stabilize these triplexes is not surprising as it has previously been shown to be selective for triplexes containing TAT triplets [25]. The triplex formed with these oligonucleotides do not contain any adjacent TAT triplets and every potential intercalation site contains protonated cytosines on either the 3′- or 5′-side. We presume that the positive charges on the third strand cytosines hinder intercalation of the charged naphthylquinoline ring. By contrast the disubstituted anthraquinones have neutral ring systems, although they possess positively charged side groups. We suggest that this is because intercalation of the uncharged anthraquinone ring system between the adjacent protonated cytosines generates a stack of alternating charged and uncharged residues, separating the charged C+GC triplets from each other. This will result in superior triplex stability in much the same way that alternating TAT and C+GC triplets generate the most stable triplexes [30–32].

By contrast the triplex formed by oligo 2 (CTTCCTTCC) is stabilized by both ligands. This triplex contains all possible combinations of adjacent triplets. Taken together with the results discussed above, this suggests that the adjacent TAT triplets form the binding site for the naphthylquinoline ligand while the anthraquinone may be able to bind in several different locations.

Neither ligand stabilizes the triplexes formed with oligos 3 or 4 in footprinting experiments and they do not affect the melting of triplexes containing alternating TAT and C+GC triplets. These triplexes are unusually stable, as predicted by Roberts and Crothers [40]. In this case we suggest that the alternation of charged and uncharged bases generates a very stable triplex, which cannot readily be disrupted by intercalation of any ligand into its structure. This would interrupt the alternation of charged and uncharged residues and so would destabilize the structure.

The suggestion that alternating positively charged and neutral rings (whether from the bases themselves or from the stabilizing ligand) generate triplexes with maximal stability, means that both positively charged and neutral ligands will be needed to stabilize all triplexes. Ligands with uncharged rings, such as the disubstituted anthraquinones, will be best suited for stabilizing C+GC-rich triplexes while ligands bearing a charged ring, such as the naphthylquinoline will be best for stabilizing TAT-rich triplexes. It appears that regions of alternating TAT and C+GC may have maximal stability and may be difficult to stabilize by adding ligands.

Acknowledgements

This work was supported by grants from Cancer Research UK and the European Union.

Ancillary