Effect of DNA target sequence on triplex formation by oligo-2′-deoxy- and 2′-O-methylribonucleotides

Abstract

The interactions of pyrimidine deoxyribo- or 2′-O-methylribo-psoralen-conjugated, triplex-forming oligonucleotides, psTFOs, with a 17-bp env-DNA whose purine tract is 5′-AGAGAGAAAAAAGAG-3′, or an 18-bp gag-DNA whose purine tract is 5′-AGG GGGAAAGAAAAAA-3′, were studied over the pH range 6.0–7.5. The stability of the triplex formed by a deoxy-env-psTFO containing 5-methylcytosines and thymines decreased with increasing pH (T_m = 56°C at pH 6.0; 27°C at pH 7.5). Replacement of 5-methylcytosines with 8-oxo-adenines reduced the pH dependence, but lowered triplex stability. A 2′-O-methyl-env-psTFO containing uracil and cytosine did not form a triplex at pH 7.5. Surprisingly, replacement of the cytosines in this oligomer with 5-methylcytosines dramatically increased triplex stability (T_m = 25°C at pH 7.5), and even greater stability was achieved by selective replacement of uracils with thymines (T_m = 37°C at pH 7.5). Substitution of the contiguous 5-methylcytosines of the deoxy-gag-psTFO with 8-oxo-adenines significantly reduced pH dependence and increased triplex stability. In contrast to the behavior of env-specific TFOs, triplexes formed by 2′-O-methyl-gag-psTFOs did not show enhanced stability. Replacement of the 3′-terminal phosphodiester of the TFO with a methylphosphonate group significantly increased the resistance of both deoxy- and 2′-O-methyl-TFOs to degradation by 3′-exonucleases, while maintaining triplex stability.

INTRODUCTION

Triplex-forming oligonucleotides (TFOs) are designed to bind to homopurine tracts in double-stranded DNA. There has been considerable interest in TFOs and psoralen-conjugated TFOs (psTFOs) because they offer the potential of modulating gene expression (1,2) either by preventing DNA transcription (3–17) or by introducing DNA mutations or deletions (18–26). A number of triplex-binding motifs have been described (27). Oligonucleotides consisting of pyrimidine bases can be designed to bind in a parallel manner to the target purine tract by forming Hoogsteen hydrogen-bonded T·AT and C⁺·GC triads, where C⁺ can be cytosine or 5-methylcytosine. Purine-containing oligonucleotides can be designed to bind through formation of reverse Hoogsteen A·AT and G·GC triads in an antiparallel manner. Oligomers having mixtures of guanine and thymine bases have also been described. These bind in either a parallel or antiparallel manner through formation of T·AT and G·GC triads.

Pyrimidine TFOs offer a design challenge because the formation of C⁺·GC triads requires protonation of the cytosine or 5-methylcytosine of the TFO (28). This requirement can reduce the effective pH range over which triplexes can form, and this restriction can serve as an obstacle to forming triplexes at physiological pH. This problem, which becomes more severe as the number of GC base pairs in the target increases, can potentially be circumvented by replacing cytosine with base analogs that do not require protonation in order to form a triad with the target GC base pair. These include pyrimidine base analogs such as pseudoisocytosine (29–32), 4-amino-5-methyl-2,6-[1H,3H]-pyrimidone (33,34), 2-aminopyridine or 2-amino-3-methylpyridine (35), or the purine analog 8-oxo-adenine (36–46).

In this study, we examine the formation of triplexes between psTFOs and two DNA targets that have different arrangements, isolated or contiguous, of their GC base pairs. The psTFOs have either deoxyribose (deoxy-psTFOs) or 2′-O-methylribose (2′-O-methyl-psTFOs) backbones and, in addition to the standard pyrimidine bases, contain varying numbers of 8-oxo-adenines. Thermal denaturation experiments carried out between pH 6.0 and 7.5 show that for a given DNA target, the stability of the triplex depends on the backbone of the psTFO and the arrangement of 8-oxo-adenines in the psTFO. In addition, replacement of the 3′-terminal phosphodiester of the TFO with a methylphosphonate group significantly increases the resistance of the TFO to degradation by 3′-exonucleases, while maintaining triplex stability.

MATERIALS AND METHODS

The 3′-O-(β-cyanoethyl-N,N-diisopropyl)phosphoramidites of 5′-O-dimethoxytrityl-N⁴-benzoyl-5-methyl-2′-deoxycyctidine, 5′-O-dimethoxytritylthymidine and 5′-O-dimethoxytrityl-N⁶-benzoyl-8-oxo-2′-deoxyadenosine, and 6-[4′-(hydroxymethyl)-4,5′,8-trimethylpsoralen]hexyl-1-O-(β-cyanoethyl-N, N-diisopropyl) phosphoramidite were purchased from Glen Research, Sterling, VA. The 3′-O-(β-cyanoethyl-N,N-diisopropyl)phosphoramidites of 5′-O-dimethoxytrityl-2′-O-methyluridine, 5′-O-dimethoxytrityl-5-methyl-2′-O-methyluridine, 5′-O-dimethoxytrityl-N⁴-acetyl-2′-O-methylcytidine, 5′-O-dimethoxytrityl-N⁴-acetyl-5-methyl-2′-O-methylcytidine, 5′-O-dimethoxytrityl-N⁶-benzoyl-8-oxo-2′-O-methyladenosine, and the 3′-O-methylphosphonamidite derivative of 5′-O-dimethoxytrityl-2′-O-methyluridine were purchased from Chem Genes Corp., Ashland, MA. The 3′-O-methylphosphonamidite derivative of 5′-O-dimethoxytritylthymidine was obtained from JBL Scientific, San Luis Obispo, CA. The nucleoside phosphoramidites were dissolved in synthesis grade acetonitrile (Fisher Scientific, Pittsburgh, PA), which was stored over calcium hydride. Reverse-phase HPLC was carried out using either a 0.46 × 15 cm Microsorb C-18 column purchased from Rainin Instruments Co., or a 0.46 × 15 cm Chrompack C-18 column purchased from Varian Instruments, Sunnyvale, CA. For analytical or preparative runs, the columns were eluted with a 20 ml linear gradient of acetonitrile in 50 mM sodium phosphate buffer pH 5.8 at a flow rate of 1.0 ml/min. Analytical runs were monitored at 260 nm and preparative runs at 290 nm. Oligomers were desalted on C-18 SEP PAK cartridges obtained from Waters, Inc., Medford, MA. The oligomers were diluted with a solution containing 2% acetonitrile in 50 mM sodium phosphate pH 5.8, and loaded onto a pre-equilibrated cartridge using a 10 ml syringe. The cartridge was washed with 10 ml of water to remove the salts and the oligomer was eluted with 3.0 ml of 50% aqueous acetonitrile. Polyacrylamide gel electrophoresis was carried out on 20 × 20 × 0.075 cm gels.

Synthesis of oligonucleotides

The sequences and designations for each oligomer are shown in Tables 1 and 2. The psTFOs were synthesized on controlled pore glass (CPG) supports using an ABI model 392 DNA/RNA synthesizer essentially as previously described (46). The protected nucleoside phosphoramidites and methylphosphonamidites were dissolved in anhydrous acetonitrile to a concentration of 0.15 M, and the nucleoside methylphosphonamidite solutions were stored for 2 h over 4 A° molecular sieves prior to use. In the case of the methylphosphonate oligomers, a 0.25 M solution of dicyanoimidazole in acetonitrile was used as the activating agent; the capping reagent B was a solution of 0.5 M dimethylaminopyridine in anhydrous pyridine; and the oxidizer contained 1.27 g of iodine, 37.5 ml of tetrahydrofuran, 12.5 ml of 2,6-lutidine and 100 µl of water. All the other oligomers were prepared using a solution of 0.45 M tetrazole in acetonitrile as activator and standard capping and oxidizer reagents purchased from Glen Research. Following the coupling step, which was 120 s for the deoxy-TFOs and 360 s for the 2′-O-methyl-TFOs, the synthesizer carried out a capping step, an oxidation step and then another capping step. After addition of the final nucleotide, the dimethoxytrityl group was removed by the synthesizer and the oligomer was conjugated with the psoralen phosphoramidite using a 600 s coupling step.

Table 1. Interactions of env-specific TFOs with env-DNA.

TFO	Sequencea	Tm°Cb
		pH 6.0	pH 6.5	pH 7.0	pH 7.5
E-d1	d-ps-TCTCTCTTTTTTCTC	56	56	35	27
E-d2	d-ps-TCTATCTTTTTTATC	35	31	23	22
E-d3	d-ps-TCTATATTTTTTATC	21	18	13	13
E-mr1	mr-ps-UCUCUCUUUUUUCUC	61	56	28	<0
E-mr2	mr-ps-TCTCTCTTTTTTCTC	65	56	33	<0
E-mr3	mr-ps-UCUCUCUUUUUUCUC	71	58	41	25
E-mr4	mr-ps-TCTCTCTTTTTTCTC	75	66	56	31
E-mr5	mr-ps-TCTATCTTTTTTATC	56	40	31	25
E-mr6	mr-ps-TCTCTCUUUUUTCTC	69	57	40	20
E-mr7	mr-ps-UCUCUCTTTTTUCUC	74	66	56	37
E-mr7p	mr-ps-UCUCUCTTTTTUCUpC	58	56	37	29

^aThe symbols are: d, 2′-deoxyribo-; mr, 2′-O-methylribo-; ps, trimethylpsoralen; C, 5-methylcytosine; A, 8-oxo-adenine; p, methylphosphonate.

^bThe T_m values are accurate to within ± 1°C.

Table 2. Interactions of gag-specific TFOs with gag-DNA.

TFO	Sequencea	Tm°Cb
		pH 6.0	pH 6.5	pH 7.0	pH 7.5
G-d1	d-ps-TCCCCCTTTCTTTTTT	55	27	14	<0
G-d2	d-ps-TAAAAATTTCTTTTTT	34	32	28	26
G-d3	d-ps-TCACACTTTCTTTTTT	58	58	34	20
G-d4	d-ps-TCAAACTTTCTTTTTT	54	54	39	29
G-d5	d-ps-TAACAATTTCTTTTTT	40	34	29	25
G-d6	d-ps-TCAAAATTTCTTTTTT	37	32	27	23
G-d7	d-ps-TAAAACTTTCTTTTTT	53	46	40	35
G-d7p	d-ps-TAAAACTTTCTTTTTpT	56	43	39	35
G-mr1	mr-ps-TCCCCCTTTCTTTTTT	56	27	<0	<0
G-mr2	mr-ps-TCACACTTTCTTTTTT	54	56	37	22
G-mr3	mr-ps-TCAAAATTTCTTTTTT	<0	<0	<0	<0
G-mr4	mr-ps-TAAAACTTTCTTTTTT	27	22	<0	<0

^aThe symbols are: d, 2′-deoxyribo-; mr, 2′-O-methylribo-; ps, trimethylpsoralen; C, 5-methylcytosine; A, 8-oxo-adenine; p, methylphosphonate.

^bThe T_m values are accurate within ± 1°C.

The synthesis cartridges containing the deoxyribo- and methylphosphonate TFOs were treated with a solution containing 0.84 M hydrazine in pyridine/acetic acid solution (4/1 v/v) for 40 h at room temperature. The hydrazine was flushed from the column. The support was washed with acetonitrile, dried under vacuum and transferred to a 4 ml autosampler vial fitted with a Teflon-lined cap. The support was treated with 400 µl of concentrated ammonium hydroxide for 2.5 h at room temperature. The supernatant was removed from the support, the support was washed with four, 200 µl aliquots of 50% aqueous acetonitrile, and the combined supernatant and washings were dried under vacuum. The residue was then treated with a solution containing 5 µl of water, 22.5 µl of acetonitrile, 22.5 µl of 95% ethanol and 50 µl of ethylenediamine for 6 h at room temperature. The solution was neutralized by adding 600 µl of ice-cold 2 M hydrochloric acid and the crude oligomers were desalted on a SEP PAK C-18 cartridge. In the case of the 2′-O-methyl-psTFOs, the hydrazine treatment was omitted from the deprotection steps.

The deoxy-psTFOs, 2′-O-methyl-psTFOs and methylphosphonate-derivatized psTFOs were purified on a C-18 reversed-phase column using a linear gradient of 2–30% acetonitrile in 50 mM sodium phosphate pH 5.8. The oligomers were desalted on a C-18 SEP PAK cartridge.

The oligomers were characterized by digestion with a combination of snake venom phosphodiesterase (SVPD) and calf intestinal alkaline phosphatase (CIP). A 0.1 A₂₆₀ unit sample of each oligomer was treated with 2 µg of SVPD and 5 U of CIP in 20 µl of buffer containing 10 mM Tris pH 8.1 and 2 mM magnesium chloride for 18 h at 37°C. The digests were analyzed by C-18 reversed-phase chromatography using a 20 ml linear gradient of 2–20% acetonitrile in 50 mM sodium phosphate pH 5.8. The deoxy-psTFOs and 2′-O-methyl-psTFOs were completely digested to their component nucleosides in the expected ratios. The methylphosphonate-derivatized oligomers were digested to their expected component nucleosides and dinucleoside methylphosphonates, deoxy- or 2′-O-methyl-NpN (p = methylphosphonate). The methylphosphonate oligomers, E-mr7p and G-d7p, were also analyzed by matrix-assisted laser desorption ionization time of flight (MALDI-TOF) mass spectrometry at the Johns Hopkins University School of Medicine Mass Spectrometry Facility: E-mr7p, m/z (expected, 5290; found, 5288); G-d7p, m/z (expected, 5316; found, 5317).

The oligonucleotides comprising the env- and gag-DNA targets were synthesized by standard solid phase phosphoramidite procedures. The oligomers were purified by strong anion exchange HPLC as previously described (46).

The UV extinction coefficients of the oligomers were determined at 260 nm as previously described (47): E-d1, 95 200; E-d2, 104 100; E-d3, 109 400; E-mr1, 116 000; E-mr2, 106 000; E-mr3, 104 500; E-mr4, 92 400; E-mr5, 102 000; E-mr6, 97 400; E-mr7, 100 000; G-d1, 104 000; G-d2, 113 200; G-d3, 110 000; G-d4, 119 000; G-d5, 118 000; G-d6, 120 000; G-d7, 116 000; G-mr1, 108 700; G-mr2, 111 000; G-mr3, 117 000; G-mr4, 116 000; E-mr7p, 100 000; G-d7p, 116 000; env-DNA purine strand, 143 000; env-DNA pyrimidine strand, 125 000; gag-DNA purine strand, 160 000; gag-DNA pyrimidine strand, 135 000.

Thermal denaturation

The target duplex and appropriate psTFO were evaporated to dryness in a 1.5 ml Eppendorf tube. The residue was dissolved in 1 ml of triplex buffer, containing 50 mM MOPS [3-(N-morpholino)propane sulfonic acid], 100 mM sodium chloride and 2.5 mM magnesium chloride. The final concentration of each strand of the triplex was 1 µM. The solution was heated at 65°C for 15 min and allowed to cool to room temperature. The solution was placed in a Speed Vac under vacuum for 1 min to remove dissolved air and then stored overnight at 4°C.

Denaturation experiments were carried out using a Varian Cary 3 E UV-visible spectrophotometer fitted with a thermostatted cell compartment. The samples were monitored at 260 nm and heated at a rate of 0.4°C over the temperature range 0.5 to 80°C or 90°C. Data were collected at 0.25°C intervals and plotted using SigmaPlot.

Stability in mammalian serum

Oligomers d-T₁₅, mr-UCAUUGACGCUGCGC, 5′-OH-E-mr7p and 5′-OH-G-d7p were each phosphorylated overnight at 37°C in 10 µl of solution that contained 0.5 µl of polynucleotide kinase, 260 µM [γ-³²P]ATP (specific activity, 10 Ci/mmol) 50 mM Tris pH 7.6, 10 mM magnesium chloride and 10 mM mercaptoethanol. A 9 µl aliquot of each solution was added to 140 µl of 0.1 M imidazole buffer pH 6.0, which contained 16 µl of 1 M 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride, and the solution was incubated at room temperature for 4 h. Ethylenediamine, 6.4 µl, was added and the pH of the solution was adjusted to 7.4 by addition of 13 µl of 12 M hydrochloric acid. The solution was incubated for 48 h at room temperature and, after dilution with 10 ml of 2% acetonitrile in 50 mM sodium phosphate, was loaded onto a C-18 SEP PAK cartridge. The cartridge was washed with 10 ml of 2% acetonitrile in 50 mM sodium phosphate, followed by 10 ml of 5% acetonitrile in 50 mM sodium phosphate, and finally with 10 ml of HPLC grade water. The oligomers were eluted with 3 ml of 50% aqueous acetonitrile. The 5′-aminoethylphosphoramidate-derivatized oligomers were purified by electrophoresis on a 20% polyacrylamide denaturing gel. The TFOs were extracted from the gel by incubating each gel slice with 1 ml of 20% acetonitrile in 0.1 M ammonium acetate pH 8.0 at 37°C for 48 h, and then once again for 1 h at 37°C using fresh solution. The oligomers were desalted on C-18 SEP PAK cartridges.

Each oligomer, 1.1 × 10⁴ c.p.m., was dissolved in 10 µl of RPMI cell culture medium that contained 10% fetal bovine serum, and the solutions were incubated at 37°C. Aliquots of 1 µl were withdrawn at 0, 0.5, 1, 1.5, 2.0 and 20 h and diluted into 4 µl of 90% formamide gel loading buffer. Following electrophoresis on a 20% polyacrylamide denaturing gel, the gel was visualized and quantitated by phosphorimaging.

RESULTS AND DISCUSSION

Preparation of TFOs and DNA targets

PsTFOs were designed to interact with the DNA duplexes shown in Figure 1. env-DNA is a 17-bp duplex whose sequence corresponds to nucleotides 7319–7335 in the env gene of human immunodeficiency virus (HIV) proviral DNA (48). This target consists of a 15-bp polypurine tract that contains five GC and 10 AT base pairs. Each of the GC base pairs is separated by at least one AT base pair. gag-DNA is an 18-bp duplex whose sequence corresponds to nucleotides 400–417 in the gag gene of HIV proviral DNA. In contrast to the polypurine tract of env-DNA, the 16-bp polypurine tract of gag-DNA contains five contiguous GC base pairs and one GC flanked by AT base pairs.

Figure 1.

Structures of deoxy- and 2′-O-methylribo- psoralen-conjugated triplex-forming oligonucleotides and the sequences of the env-DNA and gag-DNA target duplexes.

The general structure of the psTFOs is shown in Figure 1 and their sequences are shown in Tables 1 and 2. The oligomers, which are comprised of either 2′-deoxyribonucleotides (R = H) or 2′-O-methylribonucleotides (R = -OCH₃), are conjugated at their 5′ end through a six-carbon linker with trimethylpsoralen. The oligodeoxyribonucleotides contain combinations of thymine, 5-methylcytosine (C) and 8-oxo-adenine (A), whereas the oligo-2′-O-methylribonucleotides contain combinations of uracil, thymine, cytosine, 5-methylcytosine and 8-oxo-adenine. In addition, derivatives of oligomers E-mr7 and G-d7, E-mr7p and G-d7p were prepared, which each have a single methylphosphonate group as the 3′-terminal internucleotide linkage.

The oligonucleotides were prepared using standard solid phase phosphoramidite procedures. In the cases of E-mr7p and G-d7p, the methylphosphonate linkage was introduced using commercially available nucleoside 3′-O-methylphosphonamidite reagents (46). The trimethylpsoralen group was added to the 5′ ends of the oligomers using commercially available C₆-psoralen phosphoramidite reagent.

The exocyclic amino groups of the 5-methylcystosines of the oligodeoxyribonucleotides were protected with benzoyl protecting groups, whereas those of the cytosines and 5-methylcytosines of the oligo-2′-O-methylribonucleotides were protected with acetyl groups. The 6-amino groups of the 8-oxo-adenine residues in both types of oligomers were protected with benzoyl groups. The oligodeoxyribonucleotides and the methylphosphonate-derivatized oligonucleotides were deprotected by sequential treatment with hydrazine hydrate to remove the benzoyl groups of the 5-methylcytosines (46); brief, 2.5 h, treatment with concentrated ammonium hydroxide at room temperature to remove the phosphate cyanoethyl groups and to cleave the oligomer from the support; and ethylenediamine to cleave the benzoyl groups of 8-oxo-adenine (44). Hydrazine removal of the 5-methylcytosine benzoyl groups prevents potential subsequent transamination by ethylenediamine during removal of the 8-oxo-adenine benzoyl groups. In the case of the oligo-2′-O-methylribonucleotides, the hydrazine treatment was not required because the acetyl protecting groups of cytidine and 5-methylcytidine were removed by the ammonium hydroxide treatment. The methylphosphonate linkages, which are sensitive to prolonged exposure to ammonium hydroxide, are not cleaved under the deprotection conditions used here.

The oligonucleotides were purified by C-18 reversed-phase HPLC. This procedure takes advantage of the hydrophobic nature of the trimethylpsoralen group, which enables clean separation of the desired psoralen-conjugated oligomer from short, non-psoralen-containing, ‘failure sequences’. The purified oligomers were characterized by digestion to their component nucleosides using a combination of SVPD and CIP, followed by analysis by reversed-phase HPLC. The methylphosphonate oligomers were also characterized by MALDI-TOF mass spectrometry.

Thermal denaturation experiments

The interactions of the psTFOs with env-DNA or gag-DNA were monitored by UV thermal denaturation experiments over a pH range of 6.0–7.5. These experiments were carried out in a buffer that contained 50 mM MOPS, 100 mM sodium chloride and 2.5 mM magnesium chloride. Typical melting curves are shown in Figure 2. These consisted of either a single, sigmoidal-shaped curve, as exemplified by E-mr7 in Figure 2A, or a biphasic curve, as exemplified by G-d7 in Figure 2B. Curves that displayed a single transition indicated that the triplex melts as a single entity. In the case of some of the triplexes formed at low pH between the 2′-O-methyl-psTFOs and the env-DNA target, the melting temperature was higher than that of the env-DNA duplex itself (56°C). Similar results have been observed by others for triplexes formed by pyrimidine 2′-O-methyl-TFOs (49–51). In the cases where biphasic curves were observed, the first, lower temperature transition was assigned to dissociation of the third strand from the duplex. The second transition corresponds to dissociation of the target duplex. Consistent with this interpretation is the observation that the midpoint of this second transition corresponds to the melting temperature (T_m) of the target duplex in the absence of the psTFO.

Figure 2.

Thermal denaturation of (A) the E-mr7/env-DNA and E-mr7p/env-DNA triplexes, and (B) the G-d7/gag-DNA and G-d7p/gag-DNA triplexes. The experiments were carried out in a buffer containing 100 mM sodium chloride, 2.5 mM magnesium chloride and 50 mM MOPS pH 7.0.

Triplexes formed with env-DNA

The interactions of the deoxy-psTFOs and 2′-O-methyl-psTFOs with the env-DNA target are summarized in Table 1 and Figure 3A. Oligomer E-d1, which contains thymines and 5-methylcytosines, forms stable triplexes with env-DNA over the entire pH range studied. Single transition curves were observed at pH 6.0 and pH 6.5 whose T_m of 56°C coincided with the T_m of the duplex target. Biphasic transitions were observed at higher pH, and the T_m of the third strand dissociation decreased with increasing pH. The observation that essentially identical total hyperchromicities were observed at either low or high pH (data not shown) is consistent with formation of a triplex at low pH. The observed pH dependence is expected and is explained by the requirement that N-3 of the 5-methylcytosines be protonated in order to participate in triplex formation (28). Protonation enables 5-methylcytosine to donate two hydrogen bonds, one from the N⁴-exocylic amino group, the other from the N³-H, to the O⁶ and N⁷ atoms, respectively, of guanine of the target GC base pair. In addition to hydrogen bond formation, the protonated 5-methylcytosine may also help stabilize triplex formation by reducing the overall charge repulsion between the third strand and the phosphate backbones of the target duplex (52–54). Consequently, as shown in Figure 3A, the thermal stability of the third strand decreases as the pH increases, because 5-methylcytosine protonation is facilitated at low pH but becomes increasingly difficult as the pH in raised.

Figure 3.

Effect of pH on T_m by triplexes formed by (A) env-DNA and E-d1 (filled circle), E-d2 (filled square), Ed-3 (filled inverted triangle) or E-mr4 (open circle), or (B) G-d1 (filled circle), G-d2 (filled square), G-d3 (filled diamond), G-d4 (filled hexagon), G-d5 (filled cross), G-d6 (filled upright triangle) or G-d7 (filled inverted triangle).

Efforts to reduce the pH dependence of triplex formation have focused on using base analogs that do not require protonation to form two hydrogen bonds with guanine of the target GC base pair. We examined the ability of 8-oxo-adenine to facilitate triplex formation. The N⁷-H and N⁶-exocyclic amino group of this base analog are positioned to form hydrogen bonds with the N⁷ and O⁶ atoms of guanine, respectively, when the base is oriented in the syn conformation (36). Oligomers E-d2 and E-d3 contain two and three 8-oxo-adenines, respectively, in place of 5-methylcytosines in E-d1. The stabilities of the triplexes formed by either oligomer show a reduced pH dependence compared with that of the triplex formed by E-d1 (see Fig. 3A). However, the T_ms of these triplexes at low pH are significantly lower, 21–38°C, than that of the E-d1 triplex. This reduced stability may in part reflect an increased charge repulsion between the TFO and the target duplex as positively charged 5-methylcytosines are replaced with neutral 8-oxo-adenines. Consistent with this notion is the observation that the E-d3 triplex in which three 5-methylcytosines are replaced is less stable than the E-d2 triplex in which only two 5-methylcytosines are replaced.

At higher pH, protonation of the 5-methylcytosines should be considerably reduced, and consequently reduced charge repulsion should contribute less to the overall stability of the triplex. Thus one might have expected that 8-oxo-adenine substitution would serve to increase the stability of the third strand at this pH. The observation that the stabilities of the E-d2 and E-d3 triplexes are also less than of the E-d1 triplex at pH 7.5 suggests that factors other than reduced charge repulsion may contribute to triplex stabilization. It is well known that the stability of pyrimidine-type TFOs increases when 5-methylcytosines are substituted for cytosines in the TFO (55–57). Thus replacement of the 5-methylcytosines of E-d1 with cytosine essentially eliminates triplex formation at pH 7.5 (data not shown). The increased stabilization of 5-methylcytosine-substituted TFOs versus cytosine- substituted TFOs most probably results from the greater stacking interactions of the 5-methylcytosines with neighboring thymine bases (28,58). Similar effects may come into play when individual 5-methylcytosines are replaced with 8-oxo-adenine. In this case, replacement of the 5-methylcytosine with 8-oxo-adenine would decrease stacking interactions with thymine, resulting in reduced triplex stability.

Previous experiments from a number of laboratories have shown that 2′-O-methylribo-substituted pyrimidine TFOs can form stable triplexes with double-stranded DNA targets (46,49,59,60). Triplexes formed by 2′-O-methyl-TFOs often have higher stabilities than their deoxyribonucleotide counterparts. To see if this was the case in our system, a series of 2′-O-methyl-psTFOs, E-mr1 through E-mr7, were prepared and their interactions with env-DNA were studied. Oligomer E-mr1 which contains uracil and cytosine residues formed a stable triplex over the pH range 6.0–7.0 but, unlike E-d1, failed to form a triplex at pH 7.5 (see Table 1). Replacement of the uracils in E-mr1 with 5-methyluracil (thymine) to give E-mr2 should enhance stacking interactions, especially between the six contiguous thymine residues of oligomer E-mr2, and thus increase triplex stability (61,62). This was indeed the case, although only a modest increase in T_m (5°C) was observed at pH 7.0, and again no triplex formation was observed at pH 7.5.

It seemed likely that the failure of E-mr1 and E-mr2 to form triplexes at pH 7.5 was due to in large measure to the presence of the cytosine residues in these oligomers. Previous studies have suggested that contrary to the effects seen in deoxy-TFOs, substitution of cytosines with 5-methylcytosines in RNA- or 2′-O-methyl-TFOs reduces the stability of the triplex (61,62). The reason for this reduced stability is unclear. These results notwithstanding, we observed a remarkable enhancement in triplex stability when the cytosine residues of E-mr1 were replaced with 5-methylcytosine to give E-mr3. Thus stable triplex formation was observed over the entire pH range, and the T_m of the E-mr3 triplex was 25°C at pH 7.5. Further enhancement of triplex stability was effected by replacing the uracils of E-mr3 with thymines to give E-mr4. Increases in T_m of between 4 and 15°C were observed, and the T_m of the E-mr4 triplex was 31°C at pH 7.5.

Although the E-mr5 triplex is considerably more stable than its deoxyribo- counterpart, E-d2, the T_ms of both triplexes have similar pH dependencies, as shown in Figure 3A. Previous experiments by Ushijima et al. (43) have demonstrated that replacement of cytosines by 8-oxo-adenines supports triplex formation by 2′-O-methyl-TFOs. As was seen with E-d2, replacement of two of the 5-methylcytosines in E-mr4 with 8-oxo-adenine lowered triplex stability over the entire pH range. Interestingly, the T_ms of both the E-d2 and E-mr5 triplexes are decreased to the same extent. Thus over the pH range 6.0–7.0, the T_m of the E-d2 triplex is 21–25°C lower than that of E-d1 triplex, and the T_m of the E-mr5 triplex is 19–26°C lower than that of E-mr4 triplex. At pH 7.5, the differences are much less, 5 and 6°C for the E-d2 and E-mr5 triplexes, respectively. Thus it appears that similar factors determine the relative stabilities of both the deoxy-psTFO- and 2′-O-methyl-psTFO-env-DNA triplexes.

We further explored the potential contribution of base stacking to triplex stabilization by replacing various thymine residues in E-mr4 with uracils. As shown in Table 1, replacement of the six contiguous thymines significantly reduced triplex stability. Thus the T_ms of the triplex formed by E-mr6 are 6–16°C lower than those of the E-mr4 triplex over the pH range 6.0–7.5. This result suggests that enhanced base stacking interactions between the thymines in this region of E-mr4 may help stabilize the triplex. On the other hand, the T_ms of the E-mr6 triplex are very similar to the T_ms of the triplex formed by oligomer E-mr3, which contains no thymine residues. This observation suggests that stacking interactions between the thymines on the 5′ side of the 5-methylcytosines of E-mr4 could play a diminished role in stabilizing the triplex. To test this, these 5-thymines were replaced with uracils to give oligomer E-mr7. As shown in Table 1, the T_ms of the E-mr7 triplex are virtually identical to those of the E-mr4 triplex over the pH range 6.0–7.0 and 6°C higher than that of the E-mr4 triplex at pH 7.5.

Triplexes formed with gag-DNA

The env-DNA target contains five GC base pairs that are isolated from one another by intervening AT base pairs. The structure of the gag-DNA target is quite different in that it contains five contiguous GC base pairs and one isolated GC base pair. The interactions of this target with deoxy-psTFOs and 2′-O-methyl-psTFOs as monitored by UV thermal denaturation experiments are summarized in Table 2 and Figure 3B. Oligomer G-d1 consists of deoxyribonucleotides and contains 5-methylcytosines to interact with the GC base pairs of the target. Although this oligomer forms a very stable triplex at pH 6.0, the stability falls dramatically as the pH is raised, and essentially no triplex formation is observed at pH 7.5. This behavior contrasts rather sharply with that seen for the E-d1/env-DNA triplex, whose stability is much less dependent upon pH (compare Fig. 3A and B) and most probably results from the presence of the five contiguous 5-methylcytosines in G-d1. It appears that the apparent pK_as of the contiguous 5-methylcytosines of G-d1 are significantly lower than those of the isolated 5-methylcytosines of E-d1. This apparent decrease is most likely a result of increasing interbase charge repulsion arising from protonation of neighboring 5-methylcytosines (53). Thus, as bases in the 5-methylcytosine tract are protonated, it becomes increasingly more difficult for the next base to acquire a proton due the presence of positively charged neighboring bases. This charge repulsion effect is less severe when the 5-methylcytosines are isolated by intervening bases, and consequently triplex stability is less sensitive to increasing pH.

Replacement of the contiguous 5-methylcytosines of G-d1 with 8-oxo-adenines to give oligomer G-d2 dramatically enhances triplex stability at pH 7.0 and 7.5 (see Table 1) and, as shown in Figure 3B, significantly reduces the effect of pH on T_m. Thus the T_m of the triplex formed with G-d2, which contains a single 5-methylcytosine, decreases by only 8°C as the pH is raised from 6.0 to 7.5. The enhanced stability at higher pH values reflects the reduced requirement to protonate multiple 5-methylcytosines.

Elimination of all but one of the 5-methylcytosines in G-d2 reduces the potential triplex-stabilizing effect resulting from protonation of the 5-methylcytosines. Oligomers G-d3 through G-d7 were synthesized to see if retaining some of the 5-methylcytosines could enhance triplex stability. Oligomer G-d3, which contains two 8-oxo-adenines alternating with three 5-methylcytosines, formed a triplex with dramatically increased stability at low pH compared with that of the G-d2 triplex, although the T_m was 6°C less than that of the G-d2 triplex at pH 7.5. The T_ms and the pH dependence of the T_ms of the G-d3 triplex, which contains four isolated 5-methylcytosines in the third strand, are almost identical to those of the triplex formed between E-d1, which contains five isolated 5-methylcytosines, and env-DNA. This observation further supports the idea that protonation of multiple isolated 5-methylcytosines occurs more readily than does protonation of multiple contiguous 5-methylcytosines.

Replacement of the middle three 5-methylcytosines of G-d1 with 8-oxo-adenines to give G-d4 further reduced the pH dependence of the triplex T_m compared with that of the G-d3 triplex. This oligomer contains three isolated 5-methylcytosines, and this arrangement, in combination with the three contiguous 8-oxo-adenines, increased the T_m by 9°C at pH 7.5 over that of the G-d3 triplex.

Further replacement of 5-methylcytosines with four 8-oxo-adenines gave unexpected results. When the 8-oxo-adenines were separated by a 5-methylcytosine, as in G-d5, or when the 5-methylcytosine occurred at the 5′ end of the 8-oxo-adenine tract, as in G-d6, the T_ms of the triplexes decreased compared with that of the G-d4 triplex. This would be expected if protonation of 5-methylcytosines helps stabilize triplex formation as explained above. In fact, the T_ms and pH dependencies of these two triplexes, which contain two isolated 5-methylcytosines in the third strand, are very similar to those of the G-d2 triplex, which contains a single 5-methylcytosine in the third strand. However, the triplex formed by G-d7, which also contains two 5-methylcytosines, behaves quite differently. Although the pH dependence of the T_ms of this triplex are almost identical to those of the G-d5 and G-d6 triplexes (see Fig. 3B), the T_m values of the G-d7 triplex are 9–19°C higher than those of these triplexes or the triplex formed by G-d2. Since the number of 5-methylcytosines is identical in G-d5, G-d6 and G-d7, and each oligomer contains one commonly positioned 5-methylcytosine, the relative arrangement of the 8-oxo-adenines and 5-methylcytosine in the 5′ region of the oligomer must influence triplex stability, possibly as a consequence of different interactions between the bases in the third strand.

Triplexes formed by 2′-O-methyl-psTFOs and the env-DNA target were in general more stable than those formed with the deoxy-psTFOs. To see if this was also the case for the gag-DNA target, 2′-O-methylribonucleotide versions of G-d1, G-d3, G-d6 and G-d7 were prepared. As shown in Table 2, oligomer G-mr1 formed a stable triplex with gag-DNA at pH 6.0 and 6.5, but not at higher pHs. However, the T_ms of the G-mr1 triplex at the low pH values were essentially identical to those of the G-d1 triplex. Similar behavior was seen with oligomer G-mr2. This oligomer, which contains two 8-oxo-adenines alternating with the three 5-methylcytosines, formed a triplex with gag-DNA whose T_ms were almost identical to those of its deoxyribonucleotide counterpart, G-d3, over the pH range 6.0–7.5. No triplex formation was observed at any of the pHs examined with G-mr3, whose 5′ region contains a single 5-methylcytosine 5′- to the four contiguous 8-oxo-adenines. Switching the 5-methylcytosine to the 3′ end of the 8-oxo-adenine tract gave G-mr4, which formed a triplex only at pH 6.0 and 7.0. The T_ms of this triplex were less than those of the G-mr1 triplex at the same pH. Thus, unlike the situation with the env-DNA target, conversion of the TFO backbone from deoxyribo- to the 2′-O-methylribo- either does not affect triplex stability or leads to loss of stability in the gag-DNA target system.

Triplex formation by methylphosphonate-substituted psTFOs

We have previously shown that deoxy-psTFOs with alternating methylphosphonate and phosphodiester linkages are resistant to hydrolysis by the 3′-exonuclease activity found in mammalian serum (46). It seemed likely that a single methylphosphonate linkage at the 3′ end of a psTFO might be sufficient to prevent or at least significantly retard 3′-exonuclease degradation. To test this, we prepared two oligomers whose sequences were identical to E-mr7 or G-d7 and whose 3′-terminal phosphodiester linkage was replaced with a methylphosphonate linkage, 5′-OH-E-mr7p and 5′-OH-G-d7p. The 5′ ends of these oligomers were phosphorylated and then derivatized with ethylenediamine to give oligomers with a 5′-³²P-labeled 2-aminoethylphosphoramidate group, ae-p-G-d7p and ae-p-E-mr7p. This derivatization was included to prevent the possible removal of the 5′ label by phosphatase activity in the serum.

Each oligomer was incubated in cell culture medium containing 10% fetal calf serum, and aliquots were withdrawn at various times for analysis by polyacrylamide gel electrophoresis. The results are shown in Figure 4. In contrast to the control oligodeoxyribonucleotide and oligo-2′-O-methylribonucleotides whose half-lives were ≤30 min, ae-p-G-d7p and ae-p-E-mr7p both remained largely intact after 20 h incubation. These results show that the 3′-terminal methylphosphonate linkage is able to protect both deoxyribo- and 2′-O-methylribo-TFOs from exonuclease degradation.

Figure 4.

Stabilities of d-T₁₅ (filled circle), mr-UCAUUGACGCUGCGC (filled inverted triangle), G-d7p (open circle) and E-mr7p (open inverted triangle) in cell culture medium containing 10% fetal calf serum.

Methylphosphonate-derivatized psTFOs, E-mr7p and G-d7p, were prepared and their interactions with env-DNA and gag-DNA were studied by UV thermal denaturation experiments. As shown in Table 1, E-mr7p formed stable triplexes with env-DNA over the entire pH range studied. However, the T_ms of the triplex were 10–19°C lower than the T_ms of the corresponding E-mr7 triplex. In contrast, as shown in Table 2, the T_ms of the triplex formed between G-d7p and gag-DNA were almost identical to those of the G-d7 triplex. The transition curves for the third strand dissociation of E-mr7p and G-d7p are rather broad (compare G-d7p with G-d7 in Fig. 2B). Both E-mr7p and G-d7p exist as a pair of diastereoisomers, and no attempts were made to separate the two species. The observed T_ms thus represent an average of the T_ms of the two individual diastereoisomers, and this most probably accounts for the broader transition curves.

The reduced T_m of E-mr7p compared with all phosphodiester E-mr7 may be a consequence of positioning the methylphosphonate linkage next to the 3′-terminal 5-methylcytosine. Nuclear magnetic resonance (NMR) studies have suggested that terminal cytosines of TFOs have a lower apparent pK_a than internal cytosines (58,63), and previous studies with methylphosphonate TFOs have shown that the apparent pK_a of cytosine in a methylphosphonate TFO is significantly lower than cytosine in an all-phosphodiester TFO of the same sequence (52). NMR studies have also suggested that triplexes melt from the ends toward the middle (64). Thus the 3′-terminal methylphosphonate/5-methylcytosine combination in E-mr7p may promote end fraying of the third strand, resulting in lower triplex stability. This problem does not occur in the case of G-d7p, which has a 3′-terminal methylphosphonate/thymine combination.

CONCLUSIONS

Our results show that triplex formation is very sensitive to the nature of the target DNA and the structure of the TFO. Although the numbers of GC base pairs of the env-DNA and gag-DNA targets are similar, the arrangement of these base pairs, isolated or contiguous, appears to dictate the type of TFO that can bind productively. In the case of the env-DNA target, which contains six alternating AT/GC base pairs at its 5′ end, the most stable triplexes were formed by the 2′-O-methylribo-TFOs. This result is consistent with other studies that have found that pyrimidine 2′-O-methylribo-TFOs generally form highly stable triplexes with DNA targets (46,49,59,60). The source of this enhanced stabilization is not clear. 2′-O-Methyloligonucleotides most probably exist in a pre-organized A-type conformation. NMR and X-ray crystallographic studies on pyrimidine·purine–pyrimidine-type triplexes suggest that, overall, the structure of the triplex lies somewhere between a canonical A-type, RNA conformation and B-type, DNA conformation (27,60,65,66). Perhaps the pre-organized structure the 2′-O-methylribo-TFOs, in contrast to the more random structure of deoxy-TFOs, more closely resembles the final conformation of the third strand in the triplex, thus facilitating binding to the DNA target.

The situation is different, however, when the GC base pairs are arranged in a contiguous manner as they are at the 5′ end of the gag-DNA target. For this target, the triplexes formed by the 2′-O-methyl-TFOs are either of the same or lower stability than those formed by the deoxy-TFOs. Possibly in this case, the 5′-patch of contiguous GC base pairs creates a local conformation which is less accessible to or compatible with the pre-organized structure of the 2′-O-methyl-psTFO.

An important feature of triplex formation by pyrimidine TFOs is their dependence on pH, which arises from the requirement for protonating the cytosines of the TFO. Although this requirement can limit the pH at which triplex formation can occur, cytosine protonation can also contribute to overall stabilization of the triplex by reducing the charge repulsion between the phosphate groups of the TFO and those of the DNA target. Replacement of cytosines with a non-proton-requiring base analog, such as 8-oxo-adenine, clearly reduces the pH dependence of triplex stability. This was seen in both the env-DNA and gag-DNA target systems. However, use of this non-protonated analog can also lead to reduced triplex stability. This problem seems to be more serious in the env-DNA system than in the gag-DNA system. Our results suggest that 8-oxo-A substitution for the isolated 5-methylcytosines of the env-targeted TFOs offers no real advantage in terms of triplex stability. Triplexes formed by either the deoxy- or 2′-O-methyl- psTFOs are less stable than their 5-methylcytosine counterparts over the entire pH range studied.

In the case of the gag-DNA target, the situation is different. Here, replacement of the 5′-5-methylcytosine tract in the deoxy-psTFO with 8-oxo-adenines confers a significant stabilizing effect on the triplex at high pH. Retaining some of the protonated 5-methylcytosines in this tract, which serves to reduce backbone charge repulsion effects, leads to even greater stabilization. These results show that triplex stability can be enhanced and fine tuned by judicious selection of the sugar–phosphate backbone and bases of the TFO, and that different arrangements of these elements should be tested when designing a TFO to target a particular homopurine tract in DNA.

The availability of nuclease-resistant derivatives would be expected to facilitate the use of TFOs and psTFOs in modulating gene expression in living cells. We have previously shown that psTFOs with chimeric methylphosphonate/phosphodiester linkages have enhanced nuclease resistance, but form triplexes of reduced stability (46). The present study shows that a single 3′-methylphosphonate linkage is sufficient to impart nuclease resistance and, depending on the sequence at the 3′ end of the TFO, maintain triplex stability. Thus TFOs with an appropriate combination of methylphosphonate linkage and base may be useful in cell culture experiments.