INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

It is known that DNA adopts various configurations depending on environmental conditions and base sequences (Wells et al., 1988; Radhakrishnan and Patel, 1994a,b; Sklenar and Feigon, 1990; Rajagopal and Feigon, 1989a, b). One of these configurations is triplex DNA, formed by the binding of a third DNA strand in the major groove of a double helix (Frank-Kamenetskii and Mirkin, 1995). If the third strand is pyrimidine-rich, it binds parallel to the purine strand of the duplex via Hoogsteen basepairing (T*AT or C⁺*GC). When cytosine is present in the third strand, these triplexes tend to be stable only at low pH because of the requirement of protonation of the cytosines of the Hoogsteen pairing strand.

It has become increasingly apparent that triplex DNA structures have important biological implications: as a major structural feature in supercoiled plasmids and chromatin (Mirkin et al., 1987; Mirkin and Frank-Kamenetskii, 1994), for selectivity of protein binding (Musso et al., 1998), in gene regulation (Maher et al., 1989; Volkmann et al., 1995; Helene, 1991; Cooney et al., 1988; Degols et al., 1994; Grigoriev et al., 1992; Maher, 1992), and for genetic manipulation (Praseuth et al., 1999). In addition, there is considerable interest in the use of various triplex binding ligands to induce, enhance, or disrupt triplex formation (Cassidy et al., 1996; Xu et al., 1997; and Vigneswaran et al., 1996). The thorough understanding of the structural features of triplex DNA is a critical piece in our overall understanding of the biological functions and the potential therapeutic uses of triplex-forming DNA.

A number of triplex structures have been investigated by nuclear magnetic resonance (NMR) spectroscopy methods (Wang et al., 1996; Radhakrishnan and Patel, 1994a,b; Carbonnaux et al., 1991; Tarkoy et al., 1998; Koshlap et al., 1997; Gilbert and Feigon, 1999; Macaya et al., 1992; Radhakrishnan et al., 1991; de los Santos et al., 1989; Bartley et al., 1997). Early studies of DNA triplexes involved the triplex formation among three separate strands of DNA. In later studies the three strands were connected by linker regions. These DNAs then formed a "paperclip" configuration, with the region of interest being the triplex region in the center.

Interestingly, recent studies investigating the dependence of triplex stability on the length of the linker region revealed the surprising result that even with no linker present, a stable triplex structure could be formed (Chin et al., 2000). In this study we investigate the structural properties of this triplex formation, in which no linker regions are present. NMR data obtained for the single-stranded DNA 5'-T₁C₂T₃C₄T₅C₆C₇T₈C₉T₁₀C₁₁T₁₂A₁₃G₁₄A₁₅G₁₆A₁₇G₁₈-3' (18-mer) suggests that it exists in a triplex configuration (Fig. 1 A). Triplex formation in this 18-mer requires sharp turns of the DNA backbone, which have been observed in several DNA hairpin structures (Chou et al., 1996, 1999a,b; Mauffrett et al., 1998; Van Dongen et al., 1997; Gallego et al., 1997; Hare and Reid, 1986; Blommers et al., 1991; Avizonis and Kearns, 1995; Mariappan et al., 1996). The triplex structure under investigation here can be thought of as being comprised of two hairpin turns, stabilized by a third strand. To date, there is little understanding of how basepairing of an additional nucleotide may stabilize the residues of a hairpin turn.

View larger version (20K): [in this window] [in a new window]   FIGURE 1   Schematic representation of two possible triad formations of the 18-mer 5'-TCTCTCCTCTCTAGAGAG-3'; (A) intramolecular or (B) intermolecular. A star denotes the potential for Hoogsteen basepairing, and a dash denotes the potential for Watson-Crick basepairing.

In this study, we find that: 1) the sequence 5'-TCTCTCCTCTCTAGAGAG-3' forms an intramolecular triplex; 2) the internal triads form typical Hoogsteen*Watson/Crick (H*WC) pairing; 3) the 3' end C₆*G₁₈-C₇ turn triad forms a modified H*WC structure; and 4) this 3' end C₆*G₁₈-C₇ turn triad is critical in stabilizing triplex formation. In addition, we have developed a model of the triplex conformation of the 18-mer including a model of a prymidine "hairpin" turn stabilized by purine hydrogen bonding. The model represents the most probable conformation of the 18-mer that is in good agreement with the experimental data.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

Oligonucleotide synthesis and purification

The 18-mer was synthesized on a DNA synthesizer (Applied Biosystems Model 391, foster City, CA) using solid-support phosphoramidite chemistry (Yuhasz et al., 1987; Atkinson and Smith, 1984). After cleavage from the solid support these deoxyoligonucleotides were purified by reverse-phase cartridges (Poly-Pak, Glen Research Corporation, Sterling, VA), using a procedure recommended by the manufacturer. The chain length and purity was verified by gel electrophoresis. The concentration of single-stranded 18-mer was determined at 260 nm, using the molar extinction coefficient of 165,680 cm¹M¹ based on the calculation of Fasman method (Fasman, 1975).

Ultraviolet (UV) thermal melting and circular dichroism (CD) spectroscopy

UV absorbance versus temperature profiles and CD spectra were performed as described previously (Chin et al., 2000).

NMR spectroscopy

The 18-mer was dissolved in 0.5 ml of H₂O (with 10% D₂O) or 100% D₂O containing 0.15 M NaCl and 0.01 M acetate buffer. The samples' pH was adjusted to 4.5. Phase-sensitive two-dimensional (2-D) nuclear Overhauser and exchange spectroscopy (NOESY), double quantum-filtered correlated spectroscopy (DQF-COSY), and total correlation spectroscopy experiments were performed on either a Bruker DRX-700 or DRX-600 spectrometer (Billerica, MA) at 1°C. For NOESY experiments the mixing time was set to 200 ms. Additional 2-D DQF-COSY experiments were also performed at 25°C. Solvent resonance signal (H₂O) was suppressed with the water suppression by gradient tailored excitation pulse sequence.

Spectral assignment was carried out by standard sequential analysis procedures (Wijmenga et al., 1993). Examination of the COSY 2-D NMR spectrum provides assignment of H5 and H6 resonances of cytosine residues and leads to the identification of their H1' proton in NOESY spectra. Analysis of the entire H1'-H6/H8 region of the NOESY spectrum provides the sequential assignment of H1', H6, and H8 resonances for all residues. Along the H6/H8 chemical shift nuclear Overhauser effects (NOEs) identify H2', H2", H3', H4', and thymine methyl resonances. These assignments are then verified by analysis of H3'-H1', H3'-H4', H2'-H2", and H2'/H2"-H3' NOEs. Inspection of NOESY spectra acquired with the DNA in H₂O provides assignment of the exchangeable imino and amino protons. For thymine, the imino has an intrabase NOE to the methyl protons. For cytosine, there is a characteristic NOE between the amino protons and the intrabase H5 and H6 protons. In addition, for Hoogsteen-paired cytosines, the imino can be identified by its intrabase NOE to the amino protons. For guanine, the imino proton is identified by its NOE to the Watson-Crick basepaired amino group. For adenine, the amino protons have a strong NOE to the Watson-Crick-paired imino resonance. The H2 resonance of adenine is assigned by its small intranucleotide NOE to H1' and to the very strong NOE to the imino proton of the Watson-Crickbasepaired nucleotide.

Distance restraints were determined by analysis of NOESY spectra acquired in both D₂O and H₂O. For nonexchangeable protons, peak intensities were converted to distances by comparison to the H2'-H2" cross-peak (1.8 Å) and grouped as small (1.8-2.4 Å), medium (1.8-3.5 Å), and large (1.8-5.0 Å). For exchangeable protons, a uniform restraint of 1.8-5.0 Å was used, because exchange with water makes peak integrals independent of proton-proton distance. Glycosidic angle torsional restraints were determined based on comparison of the H5-H6 (of cytosine) NOE intensity with the H6-H1' and H8-H1' intensities.

Molecular modeling

The DNA was built as a single continuous strand using MSI software and the Insight II program (Acelrys Inc., San Diego, CA). Na⁺ ions were placed 4 Å from each phosphorous atom. The strand was loosely folded into a right-handed parallel triplex with pyrimidine-purine-pyrimidine pairs as indicated by CD experiments (Chin et al., 2000), minimized and counterions added. This was used as the starting structure for all calculations. Other globally folded structures were evaluated for compliance with the NOE data. However, the right-handed parallel pyrimidine-purine-pyrimidine described above complied the best. All calculations were performed on a Silicon Graphics Power Challenge with the Discover program using the Amber forcefield. Calculations occur in a two-step process: 1) simulated annealing without explicit water using a distant, dependent dielectric constant of 4*r and 2) minimization. Simulated annealing includes temperature ramp to 800 K, high-temperature equilibration to generate random starting structures, ramping on of restraints and force field parameters to establish backbone conformation and base orientation, and slow cooling to 300 K. Minimization includes 1000 iterations of restrained conjugate gradient and 5000 iterations of steepest descent. The program Suppose was used for root mean square deviation (RMSD) calculations. All RMSD values are reported as all atom RMSDs to the mean structure.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

Identification of triplex formation by 1-D NMR

1-D ¹H NMR data of the 18-mer in H₂O were acquired on a 600-MHz spectrometer. Fig. 2 shows the imino proton region of the data at various pH values. At pH 5 resonance signals appear between 14 and 15 ppm, typical of H3⁺ resonances of hydrogen-bonding cytosine residues of Hoogsteen-basepairing (Wang et al., 1996; Tarkoy et al., 1998; Koshlap et al., 1997). These resonances disappear at higher pH, typical of the pH dependence of triplex formation with cytosine in the Hoogsteen-binding strand. Thus, the 1-D ¹H NMR indicates the 18-mer forms a triplex structure.

View larger version (28K): [in this window] [in a new window]   FIGURE 2   The imino proton region of the 1H NMR, acquired at 600 MHz, of the 18-mer in H2O at various pH values. Resonance assignments at pH 5 are indicated above the peaks. Resonances associated with the H3+ of Hoogsteen-bonding cytosine residues disappear at higher pH values.

Assignment

Assignment of proton resonances was accomplished by typical 2-D sequential analysis as described in the Materials and Methods section. Fig. 3 shows the H6/H8-H1' and H8/H6-H3' regions of the NOESY spectra acquired in D₂O which provided sequential assignments for the H1', H3', H6, and H8 protons. The assignment is more complicated for triplex structures because intratriad NOEs between H1' of the Hoogsteen-paired bases and the H8 of the purine is also expected. These NOEs are considered to constitute proof of triplex formation. The majority of nonexchangable protons were assigned and listed in Table 1. Assignment of exchangeable protons was accomplished using the NOESY spectra of the 18-mer in H₂O (Fig. 4). The imino to imino connectivities can be traced sequentially from residue T₃ to residue C₆ of the Hoogsteen-paired section, and cross-strand from C₆ to T₁₀ of the Watson-Crick-paired section (Fig. 4 B). Resonances associated with the imino protons of all but the T₁, C₂, and T₁₂ bases have been assigned. Assignment of H3⁺ resonances of Hoogsteen-bonded cytosines was made by their characteristic downfield chemical shifts and their characteristic NOE pattern to their own NH₂ (Fig. 4 A). The two H3⁺ resonances belong to C₄ and C₆ cytosine residues. No resonance signal could be located for the H3⁺ of the C₂ residue. This indicates that considerable fraying occurs at the 5' end, through the C₂ residue, and that the C₆ residue, suspected to be involved in a tight turn, is strongly hydrogen bonded. This would be atypical of a hairpin turn. Usually the imino resonance is not seen. The structural features leading to this are discussed later. Several resonances of the of T₁₂ and A₁₃ residues show upfield shifting. This indicates that these protons are more shielded than in standard B-DNA. This shielding is typical of residues in hairpin turn regions because of severe kinking of the backbone that places the sugar residue protons and backbone H4' and H5'/H5" protons closer to the neighboring or their own base. The H5'/H5" assignments were not made because of extreme spectral overlap. Chemical shifts are listed in Table 1.

View larger version (49K): [in this window] [in a new window]   FIGURE 3   Region of NOESY spectrum acquired at 600 MHz of the 18-mer in D2O depicting the H1'-H8/H6 and H3'-H8/H6 walk used for initial resonance assignment.

View larger version (20K): [in this window] [in a new window]   FIGURE 4   Region of NOESY spectrum acquired at 600 MHz of the 18-mer in H2O depicting (A) the H3+-NH2 connectivity and assignment of cytosines of the Hoogsteen pairing bases, and (B) imino-imino connectivities.

NOE analysis

NOE data also indicate the formation of triplex DNA by the 18-mer. Typical connectivities indicating the presence of Watson-Crickbasepairing were observed: NOE between the cytosine amino and guanine imino for G:C basepairs and between thymine imino and adenine amino and H2 proton for A:T basepairs (Fig. 4 A). In similar fashion, NOEs indicative of Hoogsteen-basepairing are present: NOE between the H3⁺ of the pymimidine and H6/H8 of the purine (Fig. 4 A). In addition, numerous imino-to-imino NOEs are present (Fig. 4 B). The complete set of NOEs determined from NOESY spectra is available as Supporting Information. Individual NOEs are discussed below. Torsional restraints for glycosidic angles were determined based on the comparison of H5-H6 NOE strength with H6-H1' and H8-H1' intensities, found to be significantly weaker than the H5-H6 NOE intensities in all cases. Thus the glycosidic angles were restricted to between 90 and 175°.

Structural calculation

A total of 279 experimentally derived distance constraints and 13 torsional angles were calculated from two different 200-ms mixing time NOESY experiments (600 and 700 MHz), and were used during all calculations. Conformation of sugar residues was left unrestrained. Calculations were performed as described in Materials and Methods. During the first phase of the calculation, high temperature structures were generated to ensure that sufficiently random starting structures are used. Ramping on of NOE constraints at high temperature followed by the ramping on of force field parameters during slow cooling results in all conformations that satisfy the experimental NMR data. Analysis of the results from initial calculations indicated correct base orientation for hydrogen bonding in a typical Hoogsteen triplex hydrogen bond formation for most of the central basepairs. Therefore, a total of 16 hydrogen bonds and their associated dihedral constraints were added to the calculation. Dihedral angle constraints for the backbone in the H*WC forming central region were restricted to all allowable angles for DNA. No backbone dihedrals were used in the turn regions. In addition, dihedral constraints were applied to maintain planarity of the bases. An iterative process of evaluation of the resulting model for protons residing within 6 Å of one another and reanalysis of the NOESY data for those NOEs was used to detect additional NOEs. The final number of NOE restraints is listed in Table 2 along with calculation results. A total of 41 structures were generated, 78% of the structures converged to a single structure with an average total atom RMSD to the mean structure of 0.646 Å.

Overall conformation

The superposition of the ten lowest energy structures (of the converging 32 final structures) is presented in Fig. 5 and a stereoview of a single structure in Fig. 6. The backbone forms an intramolecular paperclip. Residues T₈ and A₁₇, C₉ and G₁₆, T₁₀ and A₁₅, and C₁₁ and G₁₄ form Watson-Crick basepairing similar to B-DNA. Residues T₁ through T₅ lie in the major groove of the Watson-Crick double helix and residues C₂ through T₅ form typical Hoogsteen basepairing with residues T₁₄ through A₁₇.

View larger version (57K): [in this window] [in a new window]   FIGURE 5   Superimposition of the 10 low energy structures. Hydrogens have been removed to simplify the figure.

View larger version (38K): [in this window] [in a new window]   FIGURE 6   Stereoview of the triplex. Hydrogens have been removed to simplify the figure.

It should be noted that the 18-mer has the potential to form an intermolecular triplex with Watson-Crick pairing of the self-complementary C₇ through G₁₈ region with residues T₁ through C₆ binding to the duplex major groove from both ends (Fig. 1 B). The major structural difference between the intramolecular binding (Fig. 1 A) and intermolecular binding (Fig. 1 B) is that, in the intermolecular triplex, the residues T₁₂ and A₁₃ would be part of a continuous B-DNA-like duplex, whereas, in the intramolecular triplex, they would be part of a sharp turn. Three lines of evidence indicate that triplex formation is indeed intramolecular: 1) chemical shift analysis displays the presence of upfield shifting of resonances associated with T₁₂ and A₁₃ sugar protons, typical of the sugar residues of hairpin turn regions (Van Dongen et al., 1997); 2) intratriad NOEs of T₁₂ and A₁₃ do not form the pattern typically found in duplex DNA; and 3) lack of resonance associated with the imino proton of residue T₁₂ is typical of hairpin formations. The detail discussion will be given in a later section (Conformation of the T₁*A₁₃-T₁₂ turn). In addition, our CD and UV studies indicate that triplex formation is concentration-independent (Chin et al., 2000). This also suggests that the triplex formed unimolecularly.

Conformation of the central T₃*A₁₅-T₁₀, C₄*G₁₆-C₉ H*WC forming region

In our model the central H*WC binding region is comprised of the T₃*A₁₅-T₁₀ and C₄*G₁₆-C₉ residues (Fig. 7). As discussed above, this section forms a well defined region with an all atom RMSD to the mean structure of 0.523 Å for T₃*A₁₅-T₁₀ and 0.500 Å for C₄*G₁₆-C₉.

View larger version (9K): [in this window] [in a new window]   FIGURE 7   Cross-section through each triad; T1*A13-T12, C2*G14-C11, T3*A15-T10, C4*G16-C9, T5*A17-T8, and C6*G18-C7.

For the T₃*A₁₅-T₁₀ triad there are clear NOEs between T₁₀:H3 and A₁₅:H2 along with A₁₅:NH₂ indicating Watson-Crick basepairing. In addition, the NOE between T₃:H3 and A₁₅:H8 indicates Hoogsteen pairing. All three residues are planar in the final model.

For the C₄*G₁₆-C₉ triad, an NOE was observed between C₄:NH and G₁₆:H8 (Fig. 4 A) and between C₄:NH₂ and C₉:NH₂, typical for Hoogsteen basepairings. NOE was observed between C₉:NH₂ and G₁₆:H1, typical for Watson-Crick basepairing. All three residues are planar in the final model.

Conformation of the C₂*G₁₄-C₁₁, and T₅*A₁₇-T₈ H*WC forming base triads

In the final model, C₂*G₁₄-C₁₁ and T₅*A₁₇-T₈ also form typical H*WC triads (Fig. 7). This section forms a well defined region with all atom RMSD values to the mean structure of 0.569 Å for T₅*A₁₇-T₈ and 0.534 Å for C₂*G₁₄-C₁₁.

For the C₂*G₁₄-C₁₁ triad an NOE was observed between C₂:NH₂ and C₁₁:NH₂, typical of Hoogsteen basepairing. NOE was observed between C₁₁:NH₂ and G₁₄:H1, typical of Watson-Crick basepairing. In the final model, residues G₁₄ and C₁₁ are planar with residue C₂ slightly out of plane. This out of plane is most likely attributable to fraying of the 5' end as discussed below.

For the T₅*A₁₇-T₈ triad there are clear NOEs between T₈:H3 and A₁₇:H2 and between T₈:H3 and A₁₇:NH₂, indicating Watson-Crick basepairing. An NOE observed between T₅:H3 and A₁₇:H8 indicates Hoogsteen pairing. In addition, the T₅:H1' to A₁₇:H8 NOE can clearly be observed. The presence of the NOE between H1' of the Hoogsteen-binding strand and H8 of the purine strand is considered proof of triplex formation. Residues T₅ and A₁₇ are planar with T₈ slightly out of plane. This seems to be attributable to the proximity of the turn region. Numerous additional intratriad and intertriad NOEs are identified and summarized in the Supporting Information section.

Conformation of the T₁*A₁₃-T₁₂ turn

In the final model the residues T₁₂, A₁₃, and T₁ appear to form a turn involving residues T₁₂ and A₁₃ (turn associated with the Watson-Crick basepairing hairpin section) with additional loose association of a fraying residue T₁ (indicated by the narrow line width of T₁ resonances). Fig. 7 depicts a cross-section through this triad. However, a stereoview highlighting the orientation of these residues relative to the whole model can also be seen in Fig. 6. The three residues form a fairly well defined turn with an all atom RMSD of 0.796 Å. These residues do not have the typical hydrogen-bonded proton resonance of the imino protons of T₁₂ and T₁ to N1 and N7 of A₁₃, respectively, in a standard H*WC pairing pattern. In consequence, the NOE between the imino proton of T₁ to H8 of A₁₃ in Hoogsteen pairing is also missing. Furthermore, NOEs typically seen in nonturn regions of Watson-Crick pairing in Fig. 1 B such as 1) T₁₂:H3 to A₁₃:H2, 2) G₁₄:H8 to A₁₃:H2'/H2", 3) C₁₁:NH₂ to A₁₃:NH₂, 4) T₁₂:CH₃ to A₁₃:NH₂, 5) C₁₁:H2' to T₁₂:H6, and 6) C₁₁:H2" to T₁₂:CH₃, are not presented in our study. These NOEs are observed for all other residues. In fact, the NH proton of residue T₁₂ was not observed, indicating that this proton is in fast exchange with the solvent. In the final model, this NH proton is fully exposed to solvent with no hydrogen-bonding possibility. The A₁₃:H4', A₁₃:H5'/H5", and T₁₂:H2'/H2" protons show highly unusual upfield shifts, indicating the influence of ring currents from a base directly above or below the proton. This is typical of residues in a hairpin turn (Van Dongen et al., 1997). The orientation of residues T₁₂, A₁₃, and G₁₄ in the final model account for the upfield shift of these resonances. In standard B-DNA the H5'/H5" and H4' protons are oriented toward the exterior of the DNA and away from the bases. This is not the case in a tight turn were the backbone kinks and comes fairly close to the internal bases (Fig. 6). Indeed, in the final model, the base of residue A₁₃ is near the H5'/H5" of residue A₁₃ and the base of residue G₁₄ is near A₁₃:H4'. The base orientation of the T₁₂ through A₁₃ turn of the 18-mer is remarkably similar to a previously reported DNA hairpin structure with the 5' bases of the turn stacking in a continuous fashion and the 3'-loop base folded out of the plane of base stacking and into the major groove (Van Dongen et al., 1997). Thus, this end appears to be a typical hairpin turn with base fraying of the 5'-T₁ residue. The RMSD of residues T₁₂ and A₁₃ is 0.742 and 0.888 Å, respectively, compared with an RMSD of 0.531 Å for the internal triplets, indicating that this region is slightly less well defined than the rest of the model.

Conformation of the C₆*G₁₈-C₇ triad

Residues C₆, G₁₈, and C₇ comprise those residues responsible for the other turn region. These three basepairs form a well defined turn with an RMSD value of 0.860 Å (Fig. 6). The turn associated with the residues C₆*G₁₈-C₇ is quite different from that of the T₁*A₁₃-T₁₂ turn. Residue G₁₈ shows linewidths typical of a nonfraying basepair and numerous NOEs to its adjacent residue A₁₇ and to residues C₆ and C₇. This indicates that there is little base fraying at the 3' end of the triplex. Indeed, NOEs typical of a Watson-Crick and Hoogsteen triad are present between these bases, indicating that they share some similarity to a standard Watson-Crick and Hoogsteen pairing. However, the NOE profile between C₆ and G₁₈ is clearly different from that of typical Hoogsteen paired bases. NOE was observed between C₆:NH₂ and G₁₈:H8, not typical of Hoogsteen pairing. Presence of the H3⁺ resonance for residue C₆ indicates its involvement in hydrogen bonding. Based on the experimentally observed data, initial molecular modeling results indicated correct base orientation and distance for hydrogen bonding between C₆:H3⁺ and G₁₈:O6. This hydrogen bond was included in subsequent calculations. In the final model the G₁₈ and C₆ bases create a planar pairing with the C₇ base displaced below the plane (Figs. 6 and 7). This creates a very energetically favorable stacking arrangement of the 3' bases where C₆ stacks on C₇, which is stacking on T₈. This arrangement shows great similarity to previously published hairpin turns in which basepair stacking is seen on the 5' side of the hairpin turn. It is important to note, however, that in this case the T₁ through T₁₂ turn being discussed is not a hydrogen bonding hairpin but rather, a turn comprised of consecutive pyrimidine residues (Fig. 6). Thus this kind of hairpin turn may only exist when additional bases are present to stabilize it.

It is interesting, therefore, to consider just how the two stacking bases of the turn (C₆ and C₇) interact with the third base (G₁₈) to form a stable triad. As discussed above, structural analysis of the NMR data indicates hydrogen bonding between the cytosine C₆ (at 5' end of the turn) and the third-base guanine G₁₈. In addition, the final model indicates hydrogen-bonding potential between the C₇ and G₁₈ bases in typical Watson-Crick fashion, even though C₇ is displaced out of the plane of the C₆ to G₁₈ basepair. It should be noted that no hydrogen bonding between C₇ and G₁₈ was included in the calculation. There is also some hydrogen bonding possible between the C₇ base and the A₁₇ base. In addition, no glycosidic angle restraints were applied for residues C₆ and C₇ during the calculation to allow orientation of the bases by NOE distance only. Final structures, however, have glycosidic angles of 141.89° for C₆ and 112.36° for C₇, well within the range of angles indicted by the intensity of the H6-H1' NOE. The fairly high RMSD of 1.165 Å for the G₁₈ residue is attributable to variation of the backbone region of the residue, whereas the position of the base is fairly well defined.

CD and UV analysis of the stabilizing influence of end basepairs

To evaluate the contribution of the end basepairs to the stability of triplex formation, two 17-mer sequences were studied by CD and UV. One of them lacks the T base at 5'-end, C(TC)₂(CT)₃(AG)_3, (5'-C₂T₃C₄T₅C₆C₇T₈C₉T₁₀C₁₁T₁₂A₁₃G₁₄A₁₅G₁₆A₁₇G₁₈-3' or 18-mer-T₁) and the other has no G base at 3'-end, (TC)₃(CT)₃(AG)₂A, (5'-T₁C₂T₃C₄T₅C₆C₇T₈C₉T₁₀C₁₁T₁₂A₁₃G₁₄A₁₅G₁₆A₁₇-3' or18-mer-G₁₈). The expected effect of the former one is the elimination of stacking effect of T₁ to A₁₃ in the type IV A₁₃-T₁₂ turn and the latter one destroys the C₆*G₁₈-C₇ base triad by exclusion of the center G base. As shown in Fig. 8, the CD spectrum of 18-mer-T₁ resembles that of the 18-mer under the same condition (150 mM NaCl, 10 mM phosphate buffer, pH 4.5) and suggests that there is significant triplex formation by 18-mer-T₁. In contrast, the CD spectrum of 18-mer-G₁₈ at pH 4.5 lacks triplex formation. Similar results are obtained by UV melting curves. The T_m of the 18-mer is 54.0 and the 18-mer-T₁ is 59.0, although the T_m of the 18-mer-G₁₈ is 41.9. These results showed that the triad C₆*G₁₈-C₇ is a critical factor in stabilization of triplex formation of the 18-mer.

View larger version (24K): [in this window] [in a new window]   FIGURE 8   CD analysis of the 18-mer (TC)3(CT)3(AG)3 (1), 17mer C(TC)2(CT)3(AG)3 (2), and 17-mer (TC) 3 (CT) 3 (AG) 2A (3) showing stabilizing effect of the 3' end G18 residue on formation of triplex DNA of the internal residues.

	CONCLUSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSION REFERENCES

Triplex formation of 5'-TCTCTCCTCTCTAGAGAG-3' was initially detected by UV and CD analyses in our laboratory (Chin et al., 2000). In this study we have pursued a more detailed investigation of the structure of this 18-mer by 1-D and 2-D NMR.

Our NMR data support the presence of triplex formation of the 18-mer. 1-D NMR spectra showing highly downfield-shifted resonances indicate the presence of H3⁺ protons of Hoogsteen-binding cytosine residues. The pH titration data verify the pH dependence of these resonances, also typical of Hoogsteen hydrogen bonding. Imino resonance and interbase NOEs typical of triad formation are present for most of the central residues. In addition, NOEs and linewidths indicate that the end turn regions are well defined turns with some fraying of the 5' end nucleotide.

The model derived from experimentally observed NOEs and dihedral angles is a triplex with typical Watson-Crick and Hoogsteen hydrogen bonding in the central two triads. In addition, the neighboring two triads show typical H*WC basepairing, which, however, have bases which are out of plane because of proximity of the turn regions. As no linker regions exist, tight turns are apparent. These turns can be thought of as hairpin turns, each stabilized by a third nucleotide in the vicinity. The two turns are different. One is a hairpin turn created by the Watson-Crick basepairing sections with the third nucleotide (T₁) coming from the Hoogsteen-pairing portion. This turn seems to be accomplished by the displacement of the purine residue out of the plane of the adjacent bases and into the major groove. This hairpin turn shares features with one previously published DNA hairpin (Van Dongen et al., 1997). The other turn is a pyrimidine fold of the continuous pyrimidine section of the 18-mer. This turn is distinctly different from a hairpin turn. In the pyrimidine fold, there is no stabilizing hydrogen bonding between the two arms of the fold. Such a fold must be stabilized by nearby residues. In this case, it is stabilized by the purine section of the 18-mer. Interestingly, CD and UV analysis of a 17-mer lacking the 3' G₁₈ (18-mer-G₁₈) residue shows no triplex formation, indicating that triad formation occurring at the C₆*G₁₈-C₇ turn is a critical feature in stabilization of the entire triplex.