ABSTRACT
We have used DNase I footprinting to examine the association kinetics of GA-, GT- and CT-containing oligonucleotides with the target sequence (GGA)5GG. (CCT)5CC. These reactions are slow yielding bimolecular association rate constants between 50 and 2000 M-1s-1. We find that GT-containing oligonucleotides bind much faster than GA- or CT-containing third strands. In each case the observed rate constants are faster at the centre than at the edges of the target site. Although several explanations can be offered for this observation, it is consistent with a model in which triplex formation at this repetitive site is achieved via intermediate complexes in which the third strand is not properly aligned with its target and which subsequently migrate to the correct position.
Oligonucleotide-directed triple helix formation offers the possibility of designing compounds with precise DNA sequence recognition properties, which may have application as anti-gene agents (1 -4 ). The third strand oligonucleotide lies in the major groove of duplex DNA forming specific hydrogen bonds to substituents on the purine bases. Two different types of triplex have been described which differ in the orientation of the third strand. Pyrimidine-rich oligonucleotides bind parallel to the duplex purine strand forming C+[middot]GC and T[middot]AT triplets (5 -7 ), while purine-rich third strands bind in an antiparallel orientation forming G[middot]GC, A[middot]AT and T[middot]AT triplets (8 -10 ). Formation of the parallel motif usually requires conditions of low pH, necessary for protonation of the third strand cytosines.
Effective use of this strategy requires that oligonucleotides not only have high affinity for their duplex target sites but possess optimal kinetic parameters, i.e. a fast rate of association combined with a slow rate of dissociation. This would optimise the ability to compete with DNA binding proteins and maximise the time the third strand oligonucleotide is bound to its target.
Early studies showed that the rate of association of triplex DNA is about three orders of magnitude slower than that of duplex DNA, with bimolecular association rate constants of ~103 M-1s-1 (11 ,12 ). Such a slow rate of oligonucleotide association may reduce the effectiveness of the anti-gene strategy, since the third strand must compete with protein factors which bind at the same DNA target. The slow rate of association can only be overcome by increasing the oligonucleotide concentration (13 -15 ). Triplex formation is thought to occur by a nucleation-zipper mechanism in which duplex and third strand are in rapid pre-equilibrium. The formation of a nucleus of three to five correctly aligned triplets is the rate determining step, which is followed by a rapid zippering up of the remainder of the sequence generating the complete triplex. To date the kinetics of triplex formation have been examined by several techniques which have provided estimates for the bimolecular association rate constant varying between 102 and 105 M-1s-1. These studies, which have largely examined the formation of parallel triplexes, have included a restriction enzyme protection assay (13 ), thermal denaturation and renaturation analysis (16 ), a filter binding assay (17 ,18 ), UV absorbance decay and fluorescence methods (15 ,19 ), stopped-flow kinetics and biomolecular interaction analysis (surface plasmon resonance) (20 ). More recently DNase I footprinting has been used to probe the kinetics of triplex formation with unmodified (21 ) or acridine-linked oligonucleotides (22 ). Footprinting is especially useful for measuring the rate of this slow reaction since the concentration of the third strand (typically micromolar) is much greater than that of the target site (nanomolar). The conditions therefore approximate to a pseudo-first order reaction and are similar to those that would be encountered in vivo. In addition this technique visualizes triplex formation when the target site is located within a longer DNA fragment and can in theory be used to compare the intensity of bands in different portions of the target site.
In this paper we have used DNase I footprinting to examine the rate of formation of both antiparallel and parallel triplexes on a fragment containing the repetitive target site (GGA)5GG. (CCT)5CC. The formation of GT- and GA-containing antiparallel triplexes at this target site has previously been described (23 ). In these studies the reaction was visualized by comparing the footprinting patterns obtained at various times after mixing the third strand oligonucleotide with its target site.
Oligonucleotides were purchased from Oswel DNA service, and were stored at -20oC in water at a concentration of 1 mM. DNase I was purchased from Sigma and stored at -20oC at a concentration of 7200 U/ml. Reverse transcriptase was purchased from Promega; restriction enzymes were purchased from Promega, Pharmacia or New England Biolabs.
The plasmid containing the triplex target site (GGA)5GG.(CCT)5CC, cloned into the SmaI site of pUC18, was prepared as previously described (23 ). The radiolabelled DNA polylinker fragment containing this insert was prepared by digesting the plasmid with HindIII, labelling at the 3'-end with [[alpha]-32P]dATP using reverse transcriptase and digesting again with EcoRI. The labelled fragment of interest was separated from the remainder of the plasmid DNA on a 6% non-denaturing polyacrylamide gel, eluted and dissolved in 10 mM Tris-HCl pH 7.5 containing 0.1 mM EDTA at a concentration of ~10 c.p.s./[mu]l (~10 nM). This procedure labelled the purine-containing strand of the triplex target site.
Association reactions were initiated by mixing the oligonucleotides with an equal volume of radiolabelled DNA (yielding a final concentration of 1-100 [mu]M). Oligonucleotides were dissolved in either 50 mM sodium acetate pH 5.0, containing 5 mM MgCl2 (for parallel triplex formation) or 10 mM Tris-HCl pH 7.5 containing 5 mM MnCl2 (for antiparallel triplex formation). G-rich oligonucleotides were first heated at 65oC for 5 min to remove any unusual oligonucleotide structures, before incubating the complexes at 20oC. The association reaction was followed by removing 3 [mu]l aliquots at various times and digesting with 1.5 [mu]l DNase I (5 U/ml for antiparallel triplexes at pH 7.5 and 10 U/ml for parallel triplexes at pH 5.0). DNase I was diluted in 20 mM NaCl containing 2 mM MgCl2 and 2 mM MnCl2. The reaction was stopped after 20 s by adding 4 [mu]l formamide containing 10 mM EDTA. Samples were heated at 95oC for 3 min prior to electrophoresis. Under these reaction conditions the third strand oligonucleotide (1-100 [mu]M) is present in vast excess over its target site (nanomolar). As a result the concentration of third strand essentially remains constant throughout the association reaction which therefore approximates to a pseudo-first order reaction. The association of the oligonucleotide with its target site is revealed by the time-dependent appearance of the DNase I footprint.
The products of DNase I digestion were resolved on 40 cm long, 0.3 mm thick, 9% polyacrylamide gels containing 8 M urea, which were run at 1500 V for ~2 h. The gels were then fixed in 10% (v/v) acetic acid, dried under vacuum at 80oC and exposed to autoradiography at -70oC using an intensifying screen. Bands in the digest were assigned by comparison with Maxam-Gilbert markers for guanine and adenine. Gels were scanned with a Hoefer GS365W microdensitometer.
The intensity of bands within the footprint was determined by densitometry. Pseudo-first order rate constants (kobs) for each reaction were estimated by fitting single exponential curves to the plots of band intensity against time using FigP for Windows. Bimolecular rate constants (k2) were estimated from the variation of the observed rate constants with oligonucleotide concentration. In several instances these values were determined from several bands within the triplex target site.
We have compared the results for antiparallel triplexes, described above, with the formation of parallel triplexes at the same target site, using the oligonucleotide (CCT)5CC, generating a complex containing only T[middot]AT and C+[middot]GC triplets. Because of the requirement for protonation of the third strand cytosines these experiments were performed at pH 5.0. DNase I footprinting patterns showing the interaction between (CCT)5CC and the target site (GGA)5GG are shown in Figure 5 . Once again the reaction is slow and can be observed on the timescale of a footprinting experiment. Visual inspection of the pattern with 15 [mu]M oligonucleotide shows little protection 5 min after mixing with a complete footprint apparent after ~20 min. As expected for a bimolecular process the reaction is faster at higher oligonucleotide concentrations so that with 25 and 100 [mu]M oligonucleotide the footprint is complete by 10 and 3.5 min, respectively. These autoradiographs were subjected to densitometric analysis generating pseudo-first order rate constants (kobs) for the disappearance of bands within the target site. Bimolecular association rate constants were derived from these data by examining the dependence of kobs on oligonucleotide concentration. A representative example is shown in Figure 6 A in which it can be seen that the rate constant varies with oligonucleotide concentration in a linear fashion up to 50 [mu]M, though the rate is slower than anticipated at the highest oligonucleotide concentration (100 [mu]M). Bimolecular rate constants (k2), estimated from the slopes of these plots, are presented in Figure 6 B for different cleavage products within the 17 bp target site. These results again suggest that association is faster at the centre of the target than towards the edges, though the variation is less pronounced than that seen with the antiparallel complexes. Although it is not possible directly to compare the rates of parallel and antiparallel triplex as the experiments were performed under different conditions (pH 5.0 for parallel and pH 7.5 for antiparallel structures) the rate of triplex formation with these CT-containing oligonucleotides is similar to that observed with the GA-containing oligonucleotides and is slower than that with the GT-containing oligonucleotides.
The results presented in this paper confirm that the formation of both parallel and antiparallel triplexes is a slow process which can be examined on the footprinting timescale. The measured rate constants vary in a linear fashion with the oligonucleotide concentration, consistent with the suggestion that we are indeed observing a bimolecular reaction under pseudo-first order reaction conditions. One possible exception is seen at the highest oligonucleotide concentration used for parallel triplex formation (Fig. 6 A) for which the rate is only 20% greater than at 50 [mu]M. This could be due to the presence of competing oligonucleotide structures at high concentrations, but could arise from experimental error since this is the fastest and least accurate data point. Pseudo-first order conditions are likely to be similar to those found in vivo in which the third strand oligonucleotide will be present at a much higher concentration than its target site. Bimolecular association rate constants, estimated from these association profiles, are within the range 100-2000 M-1s-1, similar to those reported in other studies (13 -22 ).
The values determined for the bimolecular rate constants show that triplex formation is very slow; much slower than the rate of duplex formation. The slower nature of this process is probably due to charge repulsion between the third stand and duplex and the weaker interaction generated by Hoogsteen rather than Watson-Crick hydrogen bonds. In addition the formation of triplex DNA probably requires a conformational change in the underlying duplex which may also be a slow process. The slow rate of triplex formation may limit the use of these agents for preventing the binding of protein factors. If a third strand oligonucleotide and a protein have similar affinities for a given DNA target then, if they are present in equal quantities, the protein may bind faster. However since triplexes dissociate very slowly, once these complexes have been formed they may be effective for blocking protein binding. It is not possible directly to compare the rate of formation of parallel and antiparallel triplexes from these studies, since they were performed under different conditions (pH 5 and Mg2+ for parallel, in contrast to pH 7.5 and Mn2+ for antiparallel) which were chosen to optimise triplex formation. However the results demonstrate that GT-containing oligonucleotides bind faster than GA-containing oligonucleotides.
These footprinting experiments have also revealed several aspects of triplex formation which were not apparent in previous studies. In particular we find that bands towards the centre of the target site are attenuated faster than those towards the edges; this is most noticeable for the interaction with (GGA)5GG. This observation was unexpected and has not been suggested in any other studies on the kinetics of triplex formation. This highlights one of the important strengths of using the footprinting technique to examine kinetic parameters. Other techniques, such as filter binding or BIAcore (surface plasmon resonance), can only determine the presence or absence of a stable interaction between an oligonucleotide and its target site, and cannot detect intermediate complexes generated in the reaction pathway. We can suggest several possible mechanisms to account for this apparent difference in association rates at different positions along the target sequence. The most likely explanation is that this repetitive target sequence provides multiple nucleation sites for triplex formation, generating intermediate complexes in which the third strand is not properly aligned with its target site. If we consider random collisions which generate intermediates with eight or more stable triplets then seven possible complexes can be generated as shown in Scheme 1a. The relative occupancy of each GGA repeat in the target in the initial conditions will therefore be 4:5:6:6:5:4 as indicated. The rate of triplex formation at the centre will therefore appear to be 1.5 times that at the edges. Similarly if complexes containing 11 triplets are considered in the intermediate species then there are five potential initiation complexes with relative occupancies 3:4:5:5:4:3 as shown in Scheme 1b. These initial complexes will then migrate to the correct position, covering the entire 17 bp target, at a rate determined by their rate of dissociation, which will be fastest for the shortest triplex regions. Since the initial footprints do not extend further in either the 3' or 5' directions it seems that in these misaligned complexes the remainder of the oligonucleotide must be hanging free in solution. An alternative explanation which could account for the variations in rate constant is that triplex formation begins at the centre of the target site and slowly progresses to the ends of the target sequence. We consider this to be unlikely since, once one portion of the third strand has successfully bound in the correct position, the local concentration of the remainder of the third strand will be relatively high and would be expected to bind rapidly by a zipper mechanism. It should also be remembered that these DNase I footprints do not arise from simple steric occlusion (since the enzyme and third strand are located in different DNA grooves) but must be the result of oligonucleotide-induced changes in DNA structure and/or flexibility. The variation in apparent rate constants could therefore arise if the centre of the triplex more readily adopts a DNase I resistant conformation than the ends. Such changes in DNA conformation might be better assessed using hydroxyl radical cleavage, since every band could be analyzed rather than the strongest cleavage products in the uneven DNase I cleavage pattern. However, we find that triplex formation does not generate hydroxyl radical footprints.
Scheme 1. Slipped structures formed by GA-containing third strand oligonucleotides at the target site (GGA)5GG. (a) Structures containing eight or more stable triplets; (b) structures with 11 or more triplets. In each case the relative occupancy indicates the number of combinations which contain a triplet at each GGA repeat.
The faster binding of GT- than GA-containing oligonucleotides may be due to different secondary structures formed by the third strand oligonucleotides. In particular it has been suggested that GA-containing oligonucleotides can form duplexes and that these self-structures can compete for triplex formation (24 ). An alternative explanation is that the difference reflects the greater intrinsic stability of T[middot]AT triplets; some studies have suggested that GT-containing oligonucleotides form more stable triplexes than GA-containing species (8 ), while others suggest that A[middot]AT is more stable (25 ). In any case it seems that the stability of each base triplet is influenced by the surrounding bases and experimental conditions. It may be significant that the variation in apparent association rates within the sequence is more pronounced for GA- than GT- or CT-containing oligonucleotides. If this reflects the existence of slipped species at the repetitive target site this may suggest that stable intermediates are shorter for GA- than GT- or CT-containing oligonucleotides.
This work was supported by grants from the Cancer Research Campaign and the Medical Research Council.
REFERENCES