2′-Deoxy-2′-fluoro-β-d-arabinonucleosides and oligonucleotides (2′F-ANA): synthesis and physicochemical studies

Abstract

Recently, hybrids of RNA and d-arabinonucleic acids (ANA) as well as the 2′-deoxy-2′-fluoro-d-arabinonucleic acid analog (2′F-ANA) were shown to be substrates of RNase H. This enzyme is believed to be involved in the primary mechanism by which antisense oligonucleotides cause a reduction in target RNA levels in vivo. To gain a better understanding of the properties of arabinose based oligonucleotides, we have prepared a series of 2′F-ANA sequences of homopolymeric (A and T) and mixed base composition (A, T, G and C). UV thermal melting and circular dichroic (CD) studies were used to ascertain the thermodynamic stability and helical conformation of 2′F-ANA/RNA and 2′F-ANA/DNA hybrids. It is shown that 2′F-ANA has enhanced RNA affinity relative to that of DNA and phosphorothioate DNA. The 2′-fluoroarabino modification showed favorable pairing to single-stranded DNA also. This is in sharp contrast to ANA, which forms weak ANA/DNA hybrids at best. According to the measured thermodynamic parameters for duplex formation, the increased stability of hybrids formed by 2′F-ANA (e.g., 2′F-ANA/RNA) appears to originate from conformational pre-organization of the fluorinated sugars and a favorable enthalpy of hybridization. In addition, NMR spectroscopy revealed a five-bond coupling between the 2′F and the base protons (H6/H8) of 2′-deoxy-2′-fluoro-β-d-arabinonucleosides. This observation is suggestive of a through-space interaction between 2′F and H6/H8 atoms. CD experiments indicate that 2′F-ANA/RNA hybrids adopt an ‘A-like’ structure and show more resemblance to DNA/RNA hybrids than to the pure RNA/RNA duplex. This feature is believed to be an important factor in the mechanism that allows RNase H to discriminate between 2′F-ANA/RNA (or DNA/RNA) and RNA/RNA duplexes.

INTRODUCTION

The synthesis of oligonucleotide analogs designed to modulate gene expression at the level of mRNA translation (‘antisense’ strategy) continues to be a major area of focus in medicinal chemistry (1–3). Although a wide range of potential antisense mechanisms have been reported, many antisense oligonucleotides (AON) in clinical investigations are believed to function via RNase H-mediated degradation of target mRNAs (4,5). Phosphorothioate DNA (S-DNA), the first generation nucleic acid analogs, exhibit a relatively low binding affinity for target RNA (6,7). Nevertheless, the increased cellular uptake, nuclease resistance and ability to elicit RNase H activity render them biologically active. For example, the first S-DNA drug for the treatment of cytomegalovirus retinitis (‘Vitravene’) in HIV patients was recently admitted to the market (8).

The generation of nucleic acid analogs exhibiting improved RNA and DNA binding properties while at the same time inducing RNase H activity remains a challenge of general interest (9,10). ‘Second-’ and ‘third-generation’ nucleic acid analogs are now emerging that have nuclease resistance and extremely high affinity and sequence selectivity for RNA (9). However, none of them, apart from the so-called ‘gapmer’ oligomers, e.g., 2′OMeRNA-(DNA gap)-2′OMeRNA, are recognized and cleaved by RNase H (11,12). Activation of this enzyme seems to be critical for the successful use of antisense drugs in vivo (4,12).

Recently, we showed that arabinonucleic acid (ANA) forms hybrids with RNA that are substrates of RNase H (13). ANA differ from the ribo analog (RNA) in the altered configuration at C2′, with the 2′-hydroxyl cis-oriented to the heterocyclic base (Fig. 1). Since RNA/RNA duplexes do not serve as substrate for RNase H, our findings established that the stereochemistry at C2′ (ANA versus RNA) is a key determinant in the activation of this enzyme. ANA/RNA duplexes are less stable, as evaluated by lower T_m, than the corresponding DNA/RNA and S-DNA/RNA hybrids (13,14). Studies of molecular models suggested steric hindrance to base rotation as a result of the size and β orientation of the 2′-OH group (14). We reasoned that this conformational characteristic was responsible for the reduced stability of ANA/RNA hybrids, and thus turned our attention to other ANA derivatives.

Figure 1.

Structures of ANA and 2′F-ANA.

In the present work, we set out to extend these studies to include the synthesis and characterization of 2′F-ANA oligomers containing the four native bases (A, G, C and T). The affinity (thermodynamic parameters) of mixed base 2′F-ANA sequences for RNA and single-stranded DNA (ssDNA) targets have not been previously reported. The thermal stability and circular dichroic (CD) profiles of 2′F-ANA/RNA duplexes relative to those formed by DNA, and RNA were compared, as well as the ability of the various analogs to discriminate between base mismatch formation. Our results show that 2′F-ANA/RNA hybrids of mixed base composition adopt a global helical conformation of the A-family, with more A- than B-form character. This property is shared with the native DNA/RNA hybrid, whose CD patterns are very similar to those of 2′F-ANA/RNA. According to UV melting data, the increased stability of 2′F-ANA/RNA hybrids relative to DNA/RNA appears to originate from both conformational pre-organization of the fluorinated strand and enthalpic contributions.

MATERIALS AND METHODS

General methods

All reactions were carried out in oven-dried glassware under a nitrogen atmosphere. Dichloromethane, carbon tetrachloride, pyridine, collidine and N,N-dimethylformamide were dried over calcium hydride. Triethylamine tris(hydrofluoride), diethylaminosulfur trifluoride (DAST), chlorotrimethylsilane (TMSCl), p-anisylchlorodiphenylmethane (MMTrCl), 4-dimethylaminopyridine (4-DMAP), thymine, cytosine, benzoyl chloride and isobutyryl chloride were obtained from Aldrich. The sugar derivative 1,3,5-tri-O-benzoyl-β-d-ribose was purchased from Pfanstiehl. Thin layer chromatography (TLC) was carried out on analytical Merck silica plates (Kieselgel 60 F-254; 0.2 mm thickness). Column chromatography was performed with silica gel (40–63 micron). Compounds on TLC were visualised by UV shadowing or by dipping the plate in Mohr’s solution [2.5 g of ammonium molybdate and 1 g of ceric sulphate in 100 ml of aqueous sulfuric acid (10%); w/v] followed by heating.

NMR spectra were run in CDCl₃, acetone-d6 or DMSO-d6 on a Varian XL500 instrument; ¹H and ³¹P NMR chemical shifts (δ, p.p.m.) are relative to tetramethylsilane and 85% H₃PO₄ (external reference), respectively. ¹⁹F NMR chemical shifts (δ, p.p.m.) are relative to trifluoroacetic acid (external reference). NOESY spectra were obtained with a mixing time of 200 min, a delay time of 1 s and a sample concentration of 1 mg/ml (deuterated methanol). ¹H NMR assignments were facilitated by recording homonuclear correlated spectra (COSY). MALDI-TOF mass spectra were obtained on a Kratos Kompact-III instrument operated in a positive reflector or linear mode. The matrix was 20 mM ammonium citrate in acetonitrile:water (1:1, v/v) containing 6-aza-2-thiothymine (10 mg/ml) (14). FAB mass spectra were obtained on a Kratos MS25RFA spectrometer using 2-nitrobenzylalcohol (NBA) as the matrix. Circular dichroism spectra were obtained on a Jasco J-710 spectropolarimeter equipped with a NESLAB RTE-111 circulating bath as previously described (14).

Synthesis of 2′-deoxy-2′-fluoroarabinonucleoside 3′-O-phosphoramidite monomers

The free 2′-fluoroarabinonucleosides were prepared following slight modifications of literature procedures (15–19). As an example, a detailed procedure follows below for the synthesis of 2′-fluoroarabinothymidine 3′-phosphoramidite derivative 6 (Scheme 1). Protocols for synthesis of the other nucleoside derivatives, along with their characterization, are given in the Supplementary Material available at NAR Online.

Scheme 1. (a) DAST, CH₂Cl₂, 40°C, 16 h; (b) HBr/AcOH, 16 h; (c) bis-silyted cytosine or thymine, CCl₄, reflux, 60 h; (d) conc. NH₄OH/MeOH, 16 h; (e) MMTr-Cl, pyridine, 4-DMAP, 16 h; (f) TMS-Cl, pyridine followed by BzCl, followed by aqueaous ammonia; (g) Cl-P(OCE)N(i-Pr)₂, EtN(i-Pr)₂; THF; CEO, 2-cyanoethoxy.

2-Deoxy-2-fluoro-1,3,5-tri-O-benzoyl-α-d-arabinofuranose (1). To a stirred solution of 1,3,5-tri-O-benzoyl-α-d-ribofuranose (4.88 g, 10.8 mmol) in dichloromethane (75 ml) was added dropwise DAST (5.23 g, 32.4 mmol), and the resulting mixture allowed to stir at 40°C (16 h). The reaction was cooled and quenched with saturated aqueous NaHCO₃ (50 ml). The organic layer was first washed with water (2 × 50 ml) and then with saturated aqueous NaHCO₃ (50 ml). This was then dried over MgSO₄, filtered and evaporated. The crude product was purified by silica gel column chromatography using CHCl₃ as the solvent. Yield was 3.6 g (71%) of a yellow oil. R_f (SiO₂): 0.38 (dichloromethane); ¹H NMR (500 MHz, p.p.m., CD₂Cl₂): 8.2–7.38 (m, 15 H, Bz), 6.7 (d, 1 H, J_1,2 = 20 Hz, H1), 5.61 (dd, 1 H, H3), 5.4 (d, 1 H, J_2,F = 50 Hz, H2), 4.7 (m, 3 H, H4, H5 and H5′); FAB-MS: 464.49 (calcd. 464.44).

2-Deoxy-2-fluoro-3,5-di-O-benzoyl-α-d-arabinofuranosyl bromide (2). A solution of 2-deoxy-2-fluoro-1,3,5-tri-O-benzoyl-α-d-arabinofuranose (1) (3.6 g, 7.76 mmol) in dichloromethane (17 ml) and HBr (30% in acetic acid, 4.5 ml) were allowed to stir overnight. The reaction was washed with water (2 × 15 ml), NaHCO₃ (saturated, 2 × 15 ml), dried over MgSO₄, filtered and evaporated to yield the product (2) as a brown oil (3.4 g, 97%). R_f (SiO₂): 0.53 (dichloromethane); ¹H NMR (500 MHz, p.p.m., CD₂Cl₂): 8.2–7.34 (m, 10 H, Bz), 6.78 (d, 1 H, J_1,2 = 20 Hz, H1), 5.63 (d, 1 H, J_2,F = 50 Hz, H2), 5.4 (dd, 1 H, H3), 4.78 (m, 3 H, H4, H5 and H5′); FAB-MS: 424.53 (calcd. 424.23).

2,4-Bis-O-(trimethylsilyl)thymine. A suspension of thymine (37.7 mmol), ammonium sulfate (250 mg) and hexamethyldisilazane (75 ml) was allowed to reflux overnight. Once the reaction was clear, it was allowed to cool and the excess hexamethyldisilazane was removed by evaporation to afford the bis-silylated pyrimidine which was used for the preparation of the nucleoside.

1-(2-Deoxy-2-fluoro-3,5-di-O-benzoyl-β-d-arabinofuranosyl)-thymine (3). A mixture of 2-deoxy-2-fluoro-3,5-di-O-benzoyl-α-d-arabinofuranosyl bromide (2) (1.70 g, 4.02 mmol) and 2,4-bis-O-(trimethylsilyl)thymine (1.23 g, 4.82 mmol) in CCl₄ (20 ml) was allowed to reflux for 60 h. The reaction was quenched with methanol (10 ml) and the solid formed was filtered. Evaporation of the solution followed by flash column chromatography (chloroform/methanol, gradient 20/1 to 19/1, v/v) afforded product 3 as a white powder: 376 mg. The solid retained in the filtering funnel contained some of the desired product, and this was further washed with methanol and processed as above. This afforded an additional 489 mg of product (48%, combined yield). R_f (SiO₂): 0.49 (chloroform/methanol 9:1); ¹H NMR (500 MHz, p.p.m., DMSO-d6): 11.53 (s, NH), 8.05–7.41 (m, H6 and Bz), 6.32 (dd, H1′, J_1′,2^′ = 4 Hz, J_H1^′,F = 19 Hz), 5.70 (ddd, H3′, J_2^′,3^′ = 2 Hz, J_3^′,4′ = 4 Hz, J_H3′,F = 19 Hz), 5.54 (ddd, H2′, J_1′,2′ = 4 Hz, J_2^′,3′ = 2 Hz, J_H2^′,F = 53 Hz), 4.78–4.75 (m, H5′ and H5″), 4.59–4.65 (m, H4′), 3.25–3.16 (m, H5′ and H5″), 1.61 (s, CH₃-C5); FAB-MS: 468.53 (calcd. 468.43).

1-(2-Deoxy-2-fluoro-β-d-arabinofuranosyl)thymine (4). To a suspension of 1-(2-deoxy-2-fluoro-3,5-di-O-benzoyl-β-d-arabinofuranosyl)thymine (3) (376 mg, 0.83 mmol) in methanol (5.8 ml) was added concentrated aqueous ammonia (645 µl). The solid dissolved as the mixture was allowed to stir at room temperature (16 h). The solution was evaporated and the resulting residue loaded onto a silica gel flash column. Elution with chloroform/ethanol (gradient 1/0 to 3:1, v/v) yielded the title compound as a colorless foam: 151 mg (75%). R_f (SiO₂): 0.49 (chloroform/ethanol 3:1); ¹H NMR (500 MHz, p.p.m., CD₃OD): 7.74 (s, H6), 6.16 (dd, H1′, J_1^′,2^′ = 4 Hz, J_H1^′,F = 17 Hz), 4.99 (ddd, H2′, J_1′,2′ = 4 Hz, J₂^′_,3′ = 2.5 Hz, J_H2^′_,F = 50 Hz), 4.31 (ddd, H3′, J₂^′_,3^′ = 2.5 Hz, J_3^′,4′ = 4.5 Hz, J_H,F = 20 Hz), 3.89 (dd, H4′, J_3^′,4′ = 4.5 Hz, J_{4,5′and 5′′} = 4.5 Hz), 3.85–3.72 (m, H5′ and H5″), 1.87 (s, CH₃-C5); FAB-MS (NBA): 259.88 (calcd. 260.22).

1-[2-Deoxy-2-fluoro-5-O-(4-methoxytrityl)-β-d-arabinofuranosyl]thymine (5). 2′F-AraT (4) (230 mg, 0.62 mmol) was co-evaporated with pyridine (3 × 1 ml) and then suspended in pyridine (3 ml). To this was added MMTrCl (228 mg, 0.739 mmol) and 4-DMAP (catalytic amount), and the resulting solution mixture allowed to stir overnight. The reaction was worked up by evaporating the pyridine and taking up the residue in dichloromethane. This solution was washed with NaHCO₃ (5%, 10 ml), dried (magnesium sulfate), filtered and finally evaporated to yield the crude product. Purification by flash column chromatography (chloroform) afforded the title compound as a colorless foam: 376 mg (94%). R_f (SiO₂): 0.27 (chloroform/methanol 9:1); ¹H NMR (500 MHz, p.p.m., DMSO-d6): 11.48 (br s, H-N), 7.41–6.89 (m, MMT, H6), 6.15 (dd, H1′, J_1^′,2^′ = 4.5 Hz, J_H1^′,F = 16 Hz), 5.03 (ddd, H2′, J_1′,2′ = 4 Hz, J_2^′,3′ = 2.5 Hz, J_H2^′,F = 53 Hz), 4.27 (ddd, 1H, H3′), J_2^′,3′ = 2.5 Hz, J_3^′,4′ = 4.5 Hz, J_H3^′,F = 19 Hz), 3.94 (m, H4′), 3.73 (s, CH₃O-), 3.30–3.25 (m, H5′ and H5″), 1.57 (s, CH₃-C5); FAB-MS (NBA): 532.23 (calcd. 532.56).

1-[2-Deoxy-2-fluoro-5-O-(4-methoxytrityl)-3-O-(β-cyanoethyl-N,N-diisopropylphosphoramidite)-β-d-arabinofuranosyl]thymine (6). N,N-diisopropyl-β-cyanoethylphosphonamidic chloride (57 mg, 0.231 mmol) was slowly added to a solution of nucleoside 5 (120 mg, 0.193 mmol) and N,N-diisopropylethylamine (60 mg, 0.463 mmol) in tetrahydrofuran (1.1 ml). The reaction turned cloudy after stirring for 30 min. After 2 h, the mixture was evaporated and the residue taken up in dichloromethane. This solution was washed with 5% aq. NaHCO₃, dried (magnesium sulfate) and concentrated. The product was isolated by chromatography on a silica gel column with a solvent of hexanes/dichloromethane/triethylamine 50/45/5 v/v/v to yield the title compound as a colorless foam: 136 mg (93%). R_f (SiO₂): 0.76, 0.83 (chloroform/ethyl acetate/ethanol 10:9:1, v/v/v); ¹⁹F NMR (407.3 MHz, p.p.m., acetone-d6): –198.46, –198.63 (ddd, J_H1^′,F = 18 Hz, J_H2^′,F = 54 Hz, J_H3^′,F = 18 Hz); ³¹P NMR (202.3 MHz, p.p.m., acetone-d6): 151.27, 150.75; FAB-MS (NBA/NaCl): M + Na⁺ = 755.78 (calcd. 755.18).

Oligonucleotide synthesis

Controlled pore glass was used as the solid support, and was derivatized with the four araF nucleosides according to the procedure of Damha et al. (20). Loadings ranged from 25 to 45 µmol/g, with the pyrimidines exhibiting higher loadings than the purines.

An Applied Biosystems 381A, five-base DNA synthesizer was configured with the 2′F-arabinonucleoside phosphoramidites (0.1 M for C, A and T; 0.15 M for G, in acetonitrile). Synthesis columns with solid supports derivatized with the desired 3′-araF nucleoside (1 µmol) were installed and capped prior to chain assembly (20,21). Each chain extension (ca. 90% coupling yield) was conducted as previously described (14), except that the iodine/water oxidizating reagent consisted of 0.1 M iodine in THF/water/pyridine (7/2/1). Oligonucleotides were purified by preparative polyacrylamide gel electrophoresis (PAGE), desalted Sephadex G-25, and characterized by MALDI-TOF mass spectrometry. Oligoribonucleotides were assembled on the same instrument using silyl phosphoramidite chemistry as previously described (21). DNA and thioate-DNA oligomers were obtained commercially (Dalton Chemical Laboratories, Inc., Toronto, ON and the University of Calgary DNA Synthesis Laboratory, Calgary, ALTA).

Thermal denaturation studies

UV thermal denaturation data were obtained on a Varian Cary1 UV-VIS spectrophotometer as previously described (14,22,23). T_m values were calculated using the base-line method reported by Puglisi and Tinoco (22), and has generally an uncertainty of ± 0.5°C.

Thermodynamic parameters of complexation were extracted from the melt curves, via van’t Hoff plots assuming a two state (all-or-none) model (23). Thermodynamic data were calculated from plots of reciprocal melting temperature, 1/T_m, versus the natural logarithm of total strand concentration, ln C_T (1–100 mM concentration range), according to equation 1:

1/T_m = (ΔS° – R ln 4)/ ΔH° + (R ln C_T/ΔH°) 1

The slope of this plot (R/ΔH°) can be used to calculate ΔH° and the intercept [(ΔS° –R ln 4)/ΔH°) permits calculation of ΔS°. The standard enthalpy and entropy changes were used to calculate ΔG° by application of the Gibb’s free energy equation 2:

ΔG° = ΔH° – TΔS 2

The free energy of association, or ΔG°, was calculated at 37°C. All plots were analyzed by linear regression.

Molecular modeling

Modeling simulations on araF-A were performed in vacuo to examine the puckering profiles of the nucleoside (AMBER 4.1). This was done using the conformational search module in HyperChem with a conjugate gradient limit of 0.01 kcal/mol A. Endocyclic torsional variations were first treated systematically then simultaneously by applying the usage directed scheme with restricted ranges for ring torsion flexing in conjunction with a non-Metropolis criterion as described by Kollosvary and Guida (24). High energy structures (defined as those possessing energies of 6 kcal/mol greater than the lowest accepted conformation) as well as those with relative energy differences within 0.05 kcal/mol were discarded in the post-optimization runs. All low energy conformers that fell within the specified acceptance criteria were further refined by a second series of optimizations in which only acyclic torsional parameters were varied and a lower gradient limit applied (0.0001 kcal/mol A). This procedure thereby identified the C2′endo conformer as the most stable structure. The H8-2′F distance was found to be 2.6 Å.

RESULTS

Synthesis of 2′-deoxy-2′-fluoro-β-d-arabinonucleosides

Syntheses araF-T (15,16), araF-C (17), araF-A (18) and araF-G (19) were prepared following modifications of literature procedures. For example, araF-T was synthesized from readily available 1,3,5-tri-O-benzoyl-d-ribose as outlined in Scheme 1 and Materials and Methods. The fluorinated sugar 1 was made by treatment of the starting material with DAST in dichloromethane (25). It has been reported that DAST cannot accomplish this reaction (15); however, we and Verdine (26) find that this reaction proceeds in good yield (71%). The C1′-α-bromo sugar 2 was produced by treating 1 with HBr/acetic acid according to the procedure of Tann et al. (16). N-glycosilation of the α-halosugar 2 with 2,4-bis-O-trimethylsilylthymine in carbon tetrachloride produced almost exclusively 2′,5′-O-dibenzoyl-β-d-araF-T (3). When dichloromethane was used as solvent, a small amount of the α-anomer was isolated along with 3 (5:95 ratio, respectively, by ¹H-NMR). Treatment of compound 3 with ethanolic ammonia provided the free nucleoside araF-T (4), which was further elaborated to the desired 5′-O-methoxytrityl-3′-O-phosphoramidite (6) according to Scheme 1.

The AraF-C amidite derivative (10) was prepared by a similar procedure, with the exception that cytosine required benzoylation at N4 (27) prior to the 5′-tritylation step (Scheme 1).

Characterization of 2′-deoxy-2′-fluoro-β-d-arabinonucleosides

The structure of the isolated araF-nucleoside 3′-O-phosphoramidites was verified by ³¹P, ¹⁹F-NMR and FAB mass spectrometry. Both free nucleosides, araF-C and araF-T, were identical (¹H-NMR) to the materials reported by Howell et al. (15) and Reichman et al. (17).

Coupling constant analysis of the sugar protons via ¹H-NMR is more complex than for other sugar residues because of the presence of the 2′-fluorine atom; for example, analysis of the 6–6.1 p.p.m. region of araF-C (Fig. 2) reveals a doublet of doublet (dd) for the H1′ resonance (J_H1′-F = 20 Hz; J_{H1^′-H2^′} = 4 Hz). Caution should be exercised in the use of coupling constants to establish the anomeric configuration of 2′F-nucleosides, since electronegative atoms at C2′ are known to reduce the magnitude of J_H1^′-H2′ and J_H2′_-H3′ (28). Nevertheless, comparisons can be made between araF-T (4), araF-C (8) and analogous 2′-fluorinated β-nucleosides for which the NMR data is available. For instance, Barchi et al. have studied 2′-F-β-d-dideoxyuridines (2′F ‘up’ configuration), and a coupling constant of 3.3 Hz for H1′-H2″ was reported (29). This magnitude is similar to that observed (4 Hz) for araF-T and araF-C (Fig. 2; see also Materials and Methods). In general, ³J_1^′,2′ in β-nucleosides with a 2′-ara substituent is in the range of 3–4 Hz whereas the α-nucleosides show very small or no coupling between the 1′ and 2′ protons (29).

Figure 2.

¹H NOESY spectrum of araF-C in DMSO-d6. Arrows in the structure show the NOEs detected. The region in which the H1′/H″ and H1′/H4′ NOEs appear is shown.

The anomeric identity of β-araF-T (4) and β-araF-C (8) was unequivocally established by two-dimensional NMR (NOESY) experiments. For example, the NOESY spectrum of araF-C in DMSO-d6 displayed strong H3′↔H6 and 5′OH↔H6 NOEs, consistent with the β configuration at the glycosidic bond. Also agreeing with the β-configuration is the observation of strong NOEs between H1′ and H4′, and between H1′ and H2″ (Fig. 2). Finally, there is no ambiguity in the anomeric configuration of araF-A and araF-G (i.e., α versus β N-glycosidic bond), as they were synthesized from β-d-riboA and β-d-riboG by chemical transformations that do not affect the stereochemistry at C1′ (18,19).

The ³¹P-NMR spectrum of araF-A phosphoramidite derivative 13 in acetone-d6 displayed an interesting feature, namely, a four-bond coupling between the 3′-phosphorus and 2′-fluorine (ca. 3 Hz) (Fig. 3). This long-range coupling is also evident in the ¹⁹F-NMR spectra (Fig. 3) and likely arises from the ‘W-type’ bond arrangement between the 2′F and 3′-P atoms (30). The long-range ³¹P/¹⁹F coupling was not detected for any of the other araF amidites, indicative of conformational differences among the amidite derivatives.

Figure 3.

³¹P and ¹⁹F NMR spectra of araF-A amidite. The structure shows the ‘W’ pathway between the 2′-fluorine and the 3′-phosphorus that leads to the long range ³¹P-¹⁹F coupling detected (compare ¹⁹F spectrum to the ¹⁹F/³¹P decoupled spectrum).

Conformation of 2′-deoxy-2′-fluoro-β-d-arabinonucleosides

Molecular mechanic calculations (araF-A) taken together with the observed J_H1′,H2″ (4 Hz) and the short H1′-H4′ distance (as assessed by the strong H1′↔H4′ NOE, see Fig. 2) suggest that the sugar pucker is O4′/C2′-endo. The O4′-endo sugar conformation was identified recently by Egli and co-workers in the crystal structure of DNA with incorporated β-d-araF-T units (31), and in the constituent DNA and 2′F-ANA strands of DNA/RNA (32) and 2′F-ANA/RNA hybrids in solution (33). It should be recalled that the O4′-endo (P = 90) conformation of araF-T units in the crystal structure of DNA results from steric consequences that predominate over stereoelectronic effects (gauche effect) caused by the local environment of the 2′-fluorine (31).

NMR spectroscopy revealed a five-bond coupling (2–3 Hz) between the base protons (H8/H6) and the 2′-fluorine atom of araF-nucleosides, e.g., 8, 11 and 14. This interesting feature is well documented in the literature and appears to be a common property of 2′-araF-nucleosides with the β configuration (34). One possible explanation for this effect is that H6/H8 interacts through-space with the 2′-fluorine, as suggested by Marquez et al. (29,35) for other nucleosides with a 2′- or 3′-F in an ‘up’ configuration (Fig. 4). This phenomenon was first described by Bergstrom et al. in theoretical terms (36) and is strongly supported by other NMR data. For example, no such H8-2′F coupling was found when the fluorine atom is situated in the 2′ ‘down’ position, e.g., riboF-β-A, (37,38). Conversely, when the anomeric configuration is α (i.e., the base is ‘down’), and the fluorine is in the 2′ ‘up’ position, no H8-2′F coupling was observed (35). This long range ¹H–¹⁹F coupling (1–3 Hz) is also detected in 2′F-ANA oligonucleotide strands complexed to complementary RNA (unpublished results). The distance between the fluorine atom and H6/H8 protons, as assessed by NOE data and molecular modeling (Fig. 4), varied from 2.5 to 2.8 Å.

Figure 4.

Energy minimized structure of araF-A. (A) Space filling model. Fluorine at C2′ is shown in yellow. (B) Chemical structure of araF-A showing the proposed through-space 2′F–H8 interaction.

Duplexes of 2′F-ANA with complementary ssDNA and RNA

Thermal dissociation data for several duplex sequences are summarized in Table 1. Mixed base sequences II (pyrimidine rich) and IV (≈ 1:1 pur/pyr) are complementary to the ‘R region’ of genomic HIV-1 RNA, whereas sequences III (≈ 1:1 pur/pyr) are complementary to the mRNA that codes for the enzyme luciferase. To further delineate the properties of 2′F-ANA, additional sequences of uniformly modified oligonucleotides, series I (all pyrimidines) and V (all purines), were examined. The affinity of 2′F-ANA was compared to other uniformly substituted oligonucleotides of identical sequences, namely S-DNA, DNA, RNA and ANA. Melting data were obtained in a buffer that mimics intracellular conditions, namely 140 mM K⁺, 1 mM Mg⁺² and 5 mM Na₂HPO₄ (pH 7.2) (39).

Table 1. Melting temperatures (T_m) and %H values of duplexes of DNA, S-DNA, RNA, ANA and 2′F-ANA oligomers with complementary target RNA and DNA strands^a.

^aAqueous solutions 1–2.5 × 10^-6 M of duplex. Buffer: 140 mM KCl, 1 mM MgCl₂, 5 mM Na₂HPO₄, (pH 7.2); ^bbroad transition; n.o., transition not observed. Uncertainty in T_m values is ± 0.5°C.

Binding to ssRNA. UV-melting curves of stoichiometric (1:1) mixtures of 2′F-ANA and RNA are of sigmoidal shape indicating a co-operative melting process (Fig. 5). The T_m value of 86°C for the octadecamer hybrid (sequence IV) indicates considerable stronger pairing of 2′F-ANA with its RNA complement relative to DNA (71°C), i.e., ca. ΔT_m = +1°C/bp. The following trends were observed for the AON/RNA duplexes under discussion:

Figure 5.

Melting curves of 18-bp duplexes (sequence III, Table 1). Oligonucleotides were hybridized to (top) single stranded RNA and (bottom) ssDNA. Buffer: 140 mM KCl, 1 mM MgCl₂, 5 mM Na₂HPO₄, pH 7.2.

I (pyrimidine): 2′F-ANA (52°) > DNA (39°) > RNA (34°) > S-DNA (21°) >>> ANA (-)

II (89% pyrimidine): 2′F-ANA (64°) > RNA (54°) ≈ DNA (51°) > S-DNA (38°) > ANA (32°)

III (72% pyrimidine): 2′F-ANA (76°) ≈ RNA (76°) ≈ DNA (63°) > S-DNA (50°) ≈ ANA (47°)

IV (56% pyrimidine): 2′F-ANA (86°) > RNA (82°) > DNA (71°) > S-DNA (62°) >> ANA (44°)

V (purine): RNA (34°) = S-DNA (34°) ≈ 2′F-ANA (30°) > ANA (26°) ≈ DNA (25°)

From inspection of these data, it is evident that interaction of 2′F-ANA with RNA is strong. Moreover, T_m values of duplexes increase as the purine content of the 2′F-ANA strand increases from 0 to ca. 60%. This pairing behavior of 2′F-ANA strands with target RNA is analogous to those of the corresponding DNA and RNA strands (i.e., IV > III > II > I). It is interesting to note that the all-purine series V do not follow this trend (see Discussion).

We calculated thermodynamic data for the hybridization of 2′F-ANA, DNA and RNA from 1/T_m versus ln[c] plots (series II), and have found that the ΔG° values (and with this the melting temperatures) vary to a significant extent. Among the three duplexes investigated, the DNA/RNA hybrid is the least stable, as reflected by the lower T_m and less favorable ΔG° value (Table 2). The enthalpy of formation (ΔH°) is more favorable in the case of the RNA duplex. This is consistent with the notion that the 2′-hydroxyl groups propagate stable and conserved water networks in the RNA grooves (40). The entropy of duplex formation (ΔS°) is more favorable for 2′F-ANA and DNA. 2′F-ANA appears to stack/H-bond more efficiently than the DNA oligomer, as reflected by its significantly more negative ΔH° value (–ΔΔH = 16 kcal/mol). The thermodynamic parameters thus indicate that the increased stability seen with 2′F-ANA/RNA is due to both enthalpic and entropic contributions. This is fully consistent with the notion that the fluorinated sugars within 2′F-ANA/RNA hybrids are conformationally biased towards one ring pucker (O4′-endo) (33).

Table 2. Effect of a single base mismatch within 2′F-ANA/DNA and 2′F-ANA/RNA duplexes^a.

Antisense strand	Target DNA ΔTm (mismatch–match)	Target RNA ΔTm (mismatch–match)
DNA	–3.9	–5.4
S-DNA	–6.1	–6.0
RNA	–7.2	–9.3
2′F-ANA	–8.0	–6.6

^aAntisense strand: 5′-AGC TCC CAG GCT CAG ATC-3′; mismatched target DNA (mismatched base underlined): 3′-TCG AGG GGC CGA GTC TAG-5′; mismatched target RNA: 3′-UCG AGG GGC CGA GUC UAG-5′; buffer: 140 mM KCl, 1 mM MgCl₂, 5 mM Na₂HPO₄, pH 7.2; duplex concentration: 2.3 µM.

CD spectra of the duplexes IV are displayed in Figure 6. These spectra were recorded at 5°C where only the duplex and no single strands exist. The CD spectra of the pure RNA duplex exhibit the characteristic A-form pattern (41), i.e., a positive band at 266 nm, and negative bands at 234 and 211 nm, whereas the pure DNA duplex adopts the B-form conformation. In contrast, DNA forms a duplex with RNA which give rise to spectra that indicate a form intermediate between A and B, but closer to the A-form (i.e., ‘A-like’ conformation). This intermediate conformation appears to be important in the recognition and cleavage of DNA/RNA substrates by RNase H (42). The spectrum of the 2′F-ANA/RNA duplex (IV) bears similarities to those of the pure RNA and DNA duplexes. Note, for example, the positive peak at 265 nm, characteristic of the A-form pattern, and the negative band at 212 nm, which is much reduced in comparison with the same band in the RNA/RNA spectrum. The ANA oligomer when hybridized with RNA presents intermediate or ‘A-like’ CD patterns. The spectral differences of the ANA/RNA versus 2′F-ANA/RNA duplex are mainly located in the region around 210 nm, where the intensity of the negative Cotton effect for 2′F-ANA/RNA is significantly reduced.

Figure 6.

CD spectra of 18-bp duplexes at 5°C (sequence IV, Table 1). Buffer: 140 mM KCl, 1 mM MgCl₂, 5 mM Na₂HPO₄, pH 7.2.

Binding to ssDNA. The hybridization data summarized in Table 1 show some interesting trends. The relative thermal stability of duplexes formed between sequences I-V and target ssDNA is as follows:

I (pyrimidine): 2′F-ANA (55°) > DNA (46°) > S-DNA (28°) > RNA (25°) >>> ANA (-)

II (89% pyrimidine): 2′F-ANA (54°) > DNA (50°) > S-DNA (37°) > RNA (28°) >> ANA (<16°)

III (72% pyrimidine): 2′F-ANA (66°) > DNA (60°) ≈ RNA (58°) > S-DNA (48°) >> ANA (<25°)

IV (56% pyrimidine): 2′F-ANA (74°) > DNA (67°) ≈ RNA (66°) > S-DNA (59°) >> ANA (26°)

V (purine): 2′F-ANA (63°) > DNA (46°) ≈ ANA (45°) > S-DNA (40°) ≈ RNA (39°)

Irrespective of the pyr/pur content of the antisense oligonucleotide, 2′F-ANA forms duplexes with ssDNA that are more stable than those formed by DNA, RNA, S-DNA and ANA. The enhancement in T_m for the 2′F-ANA oligomers relative to the unmodified DNA strands ranged from 4 to 17°C, and is largest for the purine sequence V. From the inspection of the T_m data, it appears that for most of the analogs, there exists a gradual sequence composition dependence on T_m, favoring the mixed pyrimidine/purine sequence, i.e., IV > III > II > I.

In sharp contrast, while ANA (2′-OH) hybridizes to RNA, the T_m data for the ANA + DNA mixtures indicates that the stability of ANA/DNA duplexes is weak at best (Table 1). The exception appears to be the ara-A₈/poly-dT duplex (43), whose thermal stability exceeds that of S-dA₈/poly-dT and rA₈/poly-dT hybrids.

Pairing specificity

To examine the effect of base mismatches on duplex stability, we obtained UV melting curves of 2′F-ANA/DNA and 2′F-ANA/RNA hybrids containing a single A/G mismatch near the center of the helix (series IV). As can be seen from the corresponding ΔT_m values (Table 3), the melting temperature decrease for the analogs and natural sequences to a similar extent. However, while the DNA duplex showed a ΔT_m value of –3.9°C per mismatch, ΔT_m of –8.0 and –7.2°C, where observed for 2′F-ANA/DNA and RNA/DNA, respectively. From this we conclude that 2′F-ANA and RNA strands had the best discrimination towards a mismatch placed in the DNA strand. This may be due to the more ordered structure (reduced flexibility) of 2′F-ANA and RNA relative to DNA and thioate-DNA strands. A similar conclusion is drawn from the mismatch experiments with the target RNA strands (Table 3).

Table 3. Thermodynamic parameters of duplexes^a.

Duplex	Tm (°C)	–ΔH° (kJ/mol)	–ΔS° (kJ/mol.K)	–ΔG° (kJ/mol)
DNA/RNA	48.3	489.1	1.40	55.7
RNA/RNA	53.1	755.9	2.20	74.9
2′F-ANA/RNA	62.2	554.3	1.53	79.3

^aFor sequence, see sequence III, Table 1; buffer: 140 mM KCl, 1 mM MgCl₂, 5 mM Na₂HPO₄, pH 7.2.

DISCUSSION

Formation of thermally stable duplexes between the AON and RNA is considered a key feature in the design of antisense therapeutics. Ribose modifications, particularly at the 2′-position, have yielded oligonucleotide analogs with remarkable affinity to target RNA, enhanced chemical stability and nuclease resistance (for excellent reviews of this area, see 9,44). X-ray crystallographic data show that 2′-modified oligoribonucleotide analogs have C3′-endo pucker conformation (45), and as a result, they are referred to as ‘RNA mimics’ (11). While this sugar pucker promotes pre-organization of the antisense strand to a more stable A-form geometry, the resulting AON/RNA hybrids are not substrates of RNase H (a critical enzyme involved in the mechanism of action of AON). Such a lack of recognition by RNase H has been solved by the use of the so-called ‘gapmer technology’, where 2′-deoxynucleotide (B-DNA like) residues are flanked at either end of the oligonucleotide by A-like 2′-modified nucleotides (11,12).

Until recently (13), there was no known example of a uniformly 2′-modified AON that provided both significant increases in thermal stability toward complementary target RNAs, as well as efficient recognition by RNase H by virtue of the ‘A-like’ conformation of the AON/RNA hybrid. Our group in collaboration with Parniak et al. (13) demonstrated that ANA (2′OH) invokes RNase H activity when hybridized to target RNA. From CD spectral studies and simple model building we concluded that the ability of RNase H to degrade RNA in ANA/RNA hybrids resulted from (i) the similarity of structure of these hybrids to that of the normal DNA/RNA heteroduplex, and (ii) the fact that the 2′-OH substituents of the arabinose sugar ring project into the major groove of the helix, at a site where it should not interfere with the binding and catalytic process of RNase H. This was consistent with the proposed recognition mechanism that includes critical interactions of RNase H amino acid residues with both RNA and DNA hybrid strands across the minor groove (46).

ANA oligomers of mixed base composition form stable duplexes with complementary RNA, although they have lesser binding affinity compared to the natural (DNA, RNA) counterparts (13,14,43). For example, the arabino substitution destabilizes a mixed-base ANA/RNA duplex by ca. 1.0–1.5°C per bp when compared with the native DNA/RNA hybrid (Table 1). Based on model building, this destabilization was presumed to derive from steric interference by the β-C2′-OH group, which is oriented into the major groove of the helix, causing slight local deformation, e.g., unstacking (14). It was therefore reasoned that decreasing the size of the arabinose 2′-substituent would increase the thermodynamic stability of the ANA/RNA heteroduplex. This prompted us to consider 2′-deoxy-2′-fluoro-β-d-arabinonucleic acid (2′F-ANA), whose synthesis (araF-T₁₁ and araF-U₁₁) was first reported by Watanabe in 1993 (47).

From an electronic point of view, fluorine is the most electronegative element with a value of 4.0 (versus 3.5 for oxygen) based on the Pauling scale (48). Secondly, fluorine has a van der Waals radius (1.47 Å) that is intermediate between that of hydrogen (1.2 Å) and a hydroxyl group (2.8 Å). Moreover, the ability of organically bound fluorine to serve as a ‘hydroxyl mimic’ and participate in hydrogen bonding has been widely discussed (49, and references therein). In fact, in a B-DNA duplex containing 2′F-araT residues, X-ray crystallographic data revealed clathrate-like ordered water structures around the fluorine atoms (31). The ‘up’ stereochemistry at C2′ was also an important consideration since 2′-fluoro-2′-deoxynucleosides with the opposite (ribo) configuration was shown to form 2′F-RNA/RNA hybrids that were not substrates for RNase H (38). Furthermore, we envisaged that the 2′F-ara configuration would ensure a ‘DNA-like’ pucker conformation by virtue of a strong gauche interaction between the ring oxygen (O4′) and the 2′-fluorine (50). This would yield 2′F-ANA/RNA hybrids that closely mimic the native DNA/RNA structure needed for RNase H cleavage. Preliminary studies confirmed these assumptions and showed that 2′F-ANA/RNA duplexes possess thermal stabilities higher than those of the corresponding DNA/RNA and S-DNA/RNA duplexes (13).

In the present study we prepared 2′F-ANA strands containing all four natural bases, as they are more ideal models for hybridization studies and biological investigations. An indication of the stability of duplexes formed between various 2′F-ANA sequences and their target DNA and RNA was obtained from UV melting profiles (Table 1; Fig. 5). Fully modified oligonucleotides consisting of 2′-deoxy-2′-fluoro-arabinose units exhibited excellent binding affinity for RNA targets. All of the sequences studied displayed positive increases of T_m relative to the native DNA strand. The highest T_m was obtained with 2′F-ANA IV (86°C), which consists of interdispersed purines and pyrimidines (ca. 1:1 ratio). The increases in T_m per bp relative to the native DNA is +0.2–0.8°C, which are lower than for 2′F-RNA which displays ΔT_m/bp values of +1 to +2°C (38). This in agreement with the notion that the 2′-deoxy-2′-fluororibose modification significantly stabilizes the already A-characteristic of the target RNA, due to the preferred C3′-endo conformation of these nucleosides (11,38). In 2′F-ANA/RNA hybrids the fluorinated sugars are organized in the O4′-endo conformation (33) which are expected to induce conformational alterations that deviate from the more stable A-form structure, e.g. see CD spectra (Fig. 6).

Conversely, 2′F-ANA/DNA duplexes are generally more stable than 2′F-RNA/DNA duplexes (unpublished results). Recent related experiments support this result. For example, DNA chimeras incorporating a few araF-T residues greatly stabilizes binding to complementary ssDNA relative to an unmodified DNA strand (47,51), whereas the incorporation of riboF-T residues within duplex DNA destabilizes the helix (51). This can be explained by the fact that 2′F-RNA has a northern (C3′-endo) conformation which makes it less compatible with an overall B-type conformation. In 2′F-ANA, on the other hand, the sugar conformation of the 2′-fluorosugars in the double helix is more northern (O4′-endo or ‘eastern’), although still compatible with a stable B-DNA conformation. It is this intermediate adaptability that allows 2′F-ANA to be compatible with both DNA and RNA (31).

Generally, duplexes between 2′F-ANA and RNA were of equal or slightly greater stability relative to those formed between 2′F-ANA and DNA. The exception was the all purine araF-A₁₈ (Table 1). As to why araF-A₁₈ binds much more strongly to DNA (T_m 63°C) than to RNA (T_m 30°C) deserves an explanation. Egli et al. (31) have suggested that within a DNA duplex, even with the O4′-endo conformation, the 2′-F (up) could not avoid a steric clash with the purine C8-H8 bond. Therefore, in a DNA duplex, incorporation of several 2′-deoxy-2′-fluoroarabinofuranosyl purine residues should in fact cause duplex destabilization. This is contrary to what we observed for araF-A₁₈, and perhaps such a steric clash, if it existed, is more pronounced with an A conformation (araF-A₁₈/rU₁₈) rather than with a B-conformation (araF-A₁₈/dT₁₈). No such destabilization is evident for mixed base 2′F-ANA since it binds strongly to both DNA and RNA single strands. Local deformation leading to unstacking may also explain the lower affinity in an all-purine 2′F-ANA strand in pairing with RNA. In addition, it is well known that homopolymeric A/T sequences exhibit structural differences relative to oligomers of mixed base composition (52).

One other analog, namely [3.2.0]-bicyclo-ANA, was also found to exhibit enhanced affinity for both RNA and DNA (53). This is in sharp contrast to ANA (2′OH) which binds only to RNA, forming weak ANA/DNA hybrids at best. The poor binding affinity of ANA (2′OH) towards ssDNA may reflect major distortions in the normal structure of the B-helix by unfavorable steric interactions involving the arabinose 2′-OH group. For example, B-form DNA duplexes containing ara-C (2′OH) or ara-T (2′OMe) revealed small but interesting duplex changes (54,55 and reviewed in 56 and 57). For example, in the X-ray study, ara-C was seen to pair normally with dG, and the ara-C unit (C2′-endo) was conformationally superimposable with the dC residue (C2′-endo) in the control duplex (54). However, the ara-C/dG base pair was displaced slightly from the helical axis, resulting in a loss of stacking with neighboring base pairs. In addition, there were also small changes in local torsion angles as well as twisting of the phosphodiester linkage to ara-C. The NMR study revealed that the araT(2′OMe)/dA base pair was buckled and bent slightly away from the center of the helix (55). If these perturbations also existed in short (ANA)_n/(DNA)_n segments, they would explain the observed low thermal stability of ANA/DNA hybrids. AraF-nucleotide residues (araF-T), on the other hand, can be accommodated by a DNA duplex without any conformational perturbation to the overall B-geometry (31).

An alternative explanation for why ANA/DNA hybrids are weaker than 2′F-ANA/DNA hybrids is that arabinonucleosides may adopt a different pucker conformation, e.g., the north or C3′-endo pucker (58,59). This would make them more compatible for binding to RNA (A-type) rather than DNA (B-type). Clearly, high-field NMR analysis and crystallographic work on ANA/DNA and ANA/RNA are needed to gain understanding of the selectivity of ANA towards RNA over DNA.

2′F-ANA and the 2′-epimeric 2′F-RNA may be regarded as members of the increasingly large family of conformationally biased oligonucleotides that includes the conformationally restricted ‘α-bicyclo-DNA’ (60), ‘locked nucleic acids’ (61) and the recently introduced analog ‘tricyclo-DNA’ (62). As in the case of these analogs, the high stabilization of 2′F-ANA/RNA duplex may be attributed, at least in part, to the pre-organization of the sugar residues, which entropically favors double helix formation. AraF sugars are more rigid than deoxyribose (33,51), most likely as a result of the strong gauche effect between the O4′ atom and the electronegative fluorine atom. The potential hydrogen bonding interaction between H2/H8 (base protons) and the ‘up’ 2′F atom as discussed above (see Results) would provide additional pre-organization to the 2′F-ANA strands (35,36).

Recent NMR studies on 2′F-ANA complexed to RNA have shown that the 2′-deoxy-2′-fluoroarabinonucleotide residues adopt an O4′-endo conformation, producing a stable helical structure that is indeed geometrically similar to the DNA/RNA structure (33). This conformation is in essence a conformation between the A and the B DNA, as corroborated by the CD profiles shown in Figure 6. Recently, Venkateswarlu and Ferguson (63) and Egli and co-workers (64) reported the results of a computational study of 2′F-ANA/RNA and DNA/RNA hybrids and concluded that their minor groove widths are geometrically very similar. These properties are believed to be key factors in the mechanism that allows RNase H to act on DNA/RNA and 2′F-ANA/RNA substrates (32).

CONCLUSIONS

The preparation and characterization of 2′F-ANA oligomers containing all natural heterocyclic bases, reported here for the first time, has afforded molecules that display an enhanced affinity for their RNA target and, as shown recently, maintain the ability to induce RNase H (13,33). Applications of ANA analogs against cellular mRNA targets are in progress.

SUPPLEMENTARY MATERIAL

See Supplementary Material available at NAR Online.