A two B–Z junction containing DNA resolves into an all right-handed double-helix

Abstract

Natural and artificial oligonucleotides are capable of assuming many different conformations and functions. Here we present results of an NMR restrained molecular modelling study on the conformational preferences of the modified decanucleotide d(^mC1G2^mC3G4C5_LG6_L^mC7G8^mC9G10)·d(^mC11G12^mC13G14C15_LG_L16^mC17-G18^mC19G20) which contains l deoxynucleotides in its centre. This chimeric DNA was expected to form a right–left–right-handed B-type double-helix (BB*B) at low salt concentration. Actually, it matured into a fully right-handed double helix with its central C_LpG_L core forming a right-handed Z-DNA helix embedded in a B-DNA matrix (BZ*B). The interplay between base–base and base–sugar stackings within the core and its immediately adjacent residues was found to be critical in ensuring the stabilisation of the right-handed helix. The structure could serve as a model for the design of antisense oligonucleotides resistant to nucleases and capable of hybridising to natural DNAs and RNAs.

INTRODUCTION

l-deoxyribose-containing oligonucleotides could gain particular importance in the context of antisense technology because of their enhanced resistance to nucleases (1,2). However, the pairing between l- and d-DNAs encounters a major steric hindrance and l-DNAs do not, in general, recognise single-stranded natural DNAs or RNAs in either parallel or antiparallel orientations (3). The difficulty of strand association stems mainly from the left-handed nature of homochiral l-DNAs, which does not permit an effective antisense strategy (3,4). Yet, terminally l-modified chimeric oligonucleotides not only exhibit improved hybridisation with natural nucleic acids, but they retain the expected resistance to nucleases (4,5). Also, particularly interesting results are obtained when l-oligonucleotides are designed as ligands for proteins or other small molecules (2,6–8).

The conformation of chimeric oligodeoxynucleotides, where l-deoxynucleotide units are inserted into a d-DNA chain, remains rather speculative. An answer to this problem will provide important information on the manner in which these chimeras can recognise RNAs and DNAs to form stable double or triple helices. Until now, the NMR analysis has revealed the loss of stability entailed by the introduction of one or two l-deoxynucleotides within d-DNAs (9). For instance, a punctual incorporation of l-sugars into one strand of a duplex does not impair base pairing and is accompanied by relatively smaller changes in the backbone angles compared to the larger changes on the base stacking. Thus, more substantial modifications in each DNA strand are needed to assess the effect of the chirality on DNA structure. Consequently, we prepared the decanucleotide d(^mC1G2^mC3G4C5_LG6_L^mC7G8^mC9G10)·d(^mC11G12^mC13G14- C15_LGL16^mC17G18^mC19G20), where the residues of the central CpG step exhibit l-chirality and ^mC is 5-methylcytosine. The latter substitution was found to facilitate the B→Z transition by generating hydrophobic interactions between the methyl groups and the sugar C1′ and C2′H atoms (10,11). The NMR under conditions of low salt concentration, coupled to a modelling study of this decanucleotide, concludes for a right-handed double-helix (BZ*B, where Z* stands for right-handed Z-DNA), in spite of the fact that a right–left–right double-helix (BB*B, where B* stands for left-handed B-DNA) was expected. Actually, the resulting structure is stabilised by favourable base–base and sugar–base stacking interactions both in the central Z* portion and at the Z*–B junctions. The T_m value of the modified decanucleotide (67.8°C) was barely 5°C below that of its natural counterpart (73.2°C), highlighting its good stability.

MATERIALS AND METHODS

DNA synthesis

Heterochiral oligodeoxynucleotide 5′-d(^mCG^mCGC_LG _L^mCG^mCG) was synthesised using the standard β-cyanoethylphosphoramidite methodology with a Model 381A (Applied Biosystems Inc.) DNA synthesiser (10 µmol scale). Natural 2′-deoxy-d-nucleoside phosphoramidites and derivatised support were from Glen Research. 2′-Deoxy-l-nucleosides were prepared according to slightly modified procedures (3,12). After deprotection with 30% aqueous ammonia for 5 h at 55°C, the crude decanucleotide was purified by anion-exchange chromatography on a DEAE–Sephadex A25 (Pharmacia) column using triethylammonium hydrogen carbonate buffer (pH 7.5, linear gradient from 0.5 to 1.5 M) as eluent. Fractions were analysed by HPLC and those having a purity >95% were combined, co-evaporated with water several times and converted to the sodium salt with 2× Dowex 50 W (yield 15.4 mg). The final purity was >98% as determined by anion-exchange HPLC. MALDI-TOF mass spectrum (3-hydroxypilinic acid/ammonium citrate matrix, Voyager DE, Perspective Biosystems): m/z found 3085.1, calculated 3086.1.

UV-melting experiments

UV-melting experiments were carried out on a Uvikon 931 (Kontron) spectrometer. The temperature of the cells was controlled by a Huber PD 415 temperature programmer connected to a refrigerated ethylene glycol/water bath (Huber Ministat). Annealing was performed by heating the sample (7.5 µM in 100 mM NaCl, 10 mM cacodylate, pH 7) at 80°C and gradually cooling it to 5°C. Then, the temperature was increased at a rate of 20°C/h and the absorbance measured at 260 nm. Digitised absorbance and temperature values were stored in a computer for subsequent plotting and analysis.

NMR spectroscopy

NMR spectra of the heterochiral DNA were acquired on a Bruker AMX 500. The spectra were processed with FELIX (Biosym/MSI) software on Silicon Graphics stations.

The sample was dissolved in 400 µl of potassium dihydrogen phosphate-disodium hydrogen phosphate buffer containing 1 mM EDTA. The resulting solution consisted of 2 mM in duplex with an ionic strength of 0.1 and a pH of 6.9.

Phase sensitive NOESY experiments in TPPI mode (13) in H₂O buffer at three different mixing times of 120, 200 and 300 ms were acquired with 4096 complex data points in t₂ and 800 real points in t₁, with a relaxation delay of 1.5 s, at different temperatures (5–40°C).

NOESY spectra in ²H₂O buffer were recorded at five different mixing times (60, 120, 180, 240 and 300 ms) with a relaxation delay of 2 s at 20°C. 2048 complex points were acquired in t₂ and 800 real points in t₁, with a spectral width of 5 kHz.

Two-dimensional (2D) P-COSY (14) and 2D TOCSY (15) were used for the proton assignments, ¹H–³¹P experiments (16,17) were used for the phosphorus assignments, and PFG-PEP HSQC (18) for the carbon assignments (19).

NMR restraints

Interproton distance restraints for structural calculations were obtained from a series of NOESY experiments recorded at various mixing times both in H₂O and D₂O. The adequate references in both H₂O and D₂O were taken into account. Distances were classified into four categories (1.8–3, 1.8–4, 1.8–5 and 1.8–6 Å) according to the nOe values and the type of protons (exchangeable or non-exchangeable).

Determination of rotamer domains for the backbone torsion angles is made possible using a combination of ¹H and ³¹P-NMR spectroscopy which permits domain definition of α, β, γ, δ, ɛ and ζ angles as already described (20,21).

Molecular dynamics protocol

A molecular dynamics simulated annealing protocol driven by experimental restraints was used to solve the solution structure of the molecule. This protocol uses torsion-angle molecular dynamics implementation (22) in the Crystallography and NMR System (CNS) program (23). The two initial strands were used under extended forms, using the CNS protocol generate_extended.inp, to avoid bias of a predetermined form.

The refinement procedure was similar to the one used in the original work of Stein et al. (22) with minor modifications. The dynamics protocol was performed in two stages. The first stage involved high temperature molecular dynamics in torsion space. Molecules were equilibrated at 20 000 K, over 60 ps, and cooled slowly at 1000 K over 60 ps. Non-bonded interactions were described by a repulsive quartic term and the calculations were performed under nOe (150 kcal/mol/Ų) and dihedral (5 kcal/mol/degrees²) restraints. The second stage involved Cartesian molecular dynamics consisting first of a cooling step from 1000 to 300 K followed by a 15 ps duration equilibration step at 300 K. During these steps non-bonded interactions were described by Lennard–Jones potentials and electrostatic terms were turned on. At the same time, force constants on dihedral angle restraints were established at 150 kcal/mol/degrees², while nOe restraints force constants were 150 kcal/mol/Å. 2000 steps of final minimisation were then performed.

Coordinates

The atomic coordinates of the 10 best structures and the average minimised one have been deposited in the Protein Data Bank with the accession number 1FV7.

RESULTS AND DISCUSSION

Particular nOes and base pairing observed in water

The main nOes observed in H₂O are indicated in Figure 1 together with those obtained in ²H₂O.

Figure 1.

Main nOe connectivities in the central region of the l-modified decanucleotide. Arrows indicate nOes between: the non-exchangeable protons (intra-strand, black); imino and sugar/base protons (intra-strand, green); amino and base protons (inter-strand, red); imino-amino (intra base pair, blue). Also see text.

NOESY spectra in H₂O were recorded primarily to assess the base pairing in the modified decanucleotide through examination of imino and amino protons. Analysis of the fingerprint region of imino protons correlated to amino, aromatic and sugar H1′ protons is presented in Figure 2. The decanucleotide is self-complementary, leading to a duplex with perfect 2-fold symmetry. The H1 imino signals of the eight central G residues are located in the 12.8–13.4 p.p.m. region and reflect the presence of 8 bp, including the two central C_L–G_L ones. The signals from the two terminal ^mC–G base pairs, which display significant fraying at 20°C, can be seen only at lower temperatures.

Figure 2.

Expanded NOESY contour plot of the decanucleotide at 20°C in H₂O showing the region of connectivities between the imino, amino, aromatic and H1′ resonances. Vertical lines indicate the imino resonances and the names of the linked proton are indicated for each cross peak; (a) and (b) designate the important cross peaks which are discussed in the text.

The signal at 12.81 p.p.m. is assigned to the guanine H1 imino protons of the two central modified base pairs. Both the chemical shift of these protons and their nOes with the cytosine amino protons (C4H) are fully compatible with a Watson–Crick C_L–G_L pairing. The H1 imino protons of G_L6(16) are further involved in nOes with the sugar H1′ and the base H6 protons of C_L5(15) (peaks a and b in Fig. 2). The arrangement of the two l-base pairs is particular. In the standard B-DNA, the interresidue distance GH1 to CH1′/H5 is too large to show such nOe effects and a possible intra-base pair origin is not feasible either, because inside a base pair the mentioned distances are too large to account for the observed nOe intensities.

In the amino to aromatic region (not shown), the cross peak at 8.1/7.10 p.p.m. arising from a cytosine is also of great interest. Since inside a cytosine base the distance between the two amino protons and the aromatic proton is >5.3 Å, the observed cross peak could only be assigned to the interstrand nOe C₄5(15)H42–C₄15(5)H6, although the two strands are quite identical. The mentioned nOe is incompatible with a B-DNA helix, where the corresponding distance measures 6.5 Å.

Thus, NOESY experiments in water provide decisive information relative to the central C_LpG_L motif in the modified decanucleotide, especially in establishing that this motif maintains a Watson–Crick base pairing while substantially deviating from B-DNA type geometry.

Particular nOes and chemical shifts observed in ²H₂O

The 2D NOESY analysis in the ²H₂O solvent confirms the large deviations from B-DNA conformation occurring at the centre of the decanucleotide. First of all, the intrastrand G_L6(16)H8–H1′/H2′/H2″/H3′C_L5(15) nOes are lacking, and the internucleotide G_L6(16)H8–H4′ nOe is unusually strong (Fig. 3). At the same time, the classical B-DNA connectivities are present for the other steps, although the one corresponding to C7(17)H6–G6(16)H1′ appears unusually weak. Secondly, intrastrand intersugar nOes are observed between several protons of the central C_LpG_L step and its immediately adjacent residues. These include the nOes between H1′ of G4(14) and H2′/H2″/H3′ of C_L5(15), and possibly those between H2′/H2″ of G4(14) and H2′/H2″ of C_L5(15), which were more difficult to assess due to resonance overlaps. Thirdly, the intranucleotide H2′–H8 and H1′–H8 nOes important for determining the base orientation relative to the sugar are weak and very strong, respectively, suggesting a syn conformation for the G_L6(16) base. This is further confirmed by the ¹³C resonance of the C1′ sugar of G_L6(16), shifted 3–4 p.p.m. to lower field, which is typical for syn bases, as underscored particularly by the ¹³C NMR studies performed on DNA tetraplexes stabilised by guanines in both anti and syn conformations (24). Finally, there are some proton resonances showing abnormally high field shifts. These include the C_L5(15) proton resonances H2′ (1.51 p.p.m.), H4′ (3.67 p.p.m. while the other H4′ are located at ∼4.10 p.p.m.), H5″ (3.68 p.p.m.) and, above all, H5′ (2.56 p.p.m., a chemical shift larger by 1.5 p.p.m. compared to that of the other H5′); all of these assignments having been ascertained using ¹H–¹³C HSQC experiments (18,24). For instance, the ¹³C resonance of the C5′ atom is associated to the H5′ and H5″ proton resonances at 3.6 and 2.55 p.p.m. High field shifts for the cytosine H2′, H5′ and H5″ sugar protons have been seen before in cytosine–guanine Z-steps (25–28). Such effects could be caused by the particular orientation of cytosine in Z-DNA, which places the above mentioned protons directly under the ring current of the neighbouring 5′ and 3′ guanine bases.

Figure 3.

Expanded NOESY contour plot of the decanucleotide at 20°C in D₂O showing the aromatic to H1′ region. Sequential connectivities are indicated. The asterisk marks the absent connectivity, which is discussed in the text.

Structure calculations and analysis

The structure of the duplex in solution was resolved by restrained torsion angle molecular dynamics, starting with the two strands in fully extended conformation. Initially, test calculations with synthetic nOes were undertaken with d(CG)₅ in order to find the appropriate force constants for the characteristic B-DNA structure. This molecule will be presented in the following section as the reference molecule.

Molecular torsion angle dynamics was carried out 50 times using random initial velocities for each structure of the chimeric duplex, attaining the success level of 56% with an overall energy of less than –500 kcal/mol. This high rate was similar to that obtained by Stein et al. (22) for their analysis of B-DNA duplexes. The molecules with the lowest energy were averaged and data for the resulting structure were summarised in Table 1.

Table 1. Statistics of the NMR restraints and structures.

NMR restraints in the complex
Intra-residue distances	168
Inter-residue distances	186
Hydrogen bond restraints	36
Dihedral angle restraints	104
Planarity restraints	10
Total restraints	504
Restraints/residue	25.2
Structure analysis
Average deviations from ideal covalent geometry
Bond length (Å)	0.0017 ± 0.00008
Bond angles (°)	0.364 ± 0.013
Improper angles (°)	0.227 ± 0.004
NOE violations
Number (>0.2 Å)	0
r.m.s.d. of the violations (Å)	0.008 ± 0.001
Dihedral angles violations
Number (>2°)	0
r.m.s.d. of the violations (°)	0.114 ± 0.035
Statistics of superpositions
r.m.s.d. (Å) with respect to the average structure:
10 structures (16 residues)	1.31 ± 0.36
10 structures (eight central residues)	0.60 ± 0.14

The r.m.s.d. of the 10 best structures with respect to the average structure was 1.31 Å, seemingly a rather good value, given the conservative restraints used in the system (notably those on nOes), comparable to those previously reported (29). Moreover, the conformation of the four central base pairs of interest, G4C_L5G_L6C7/G14C_L15G_L16C17, appeared better resolved (0.60 ± 0.14 Å) relative to the rest of the molecule.

Calculated conformation

The average structure of the decanucleotide (Fig. 4a) was compared to the structure of the d(CG)₅ molecule calculated using synthetic data. Note the Z–Z conformation of the backbone at the central C_LpG_L step which contrasts with the smooth appearance of the B-DNA backbone in the d(CG)₅ molecule. An additional feature is the ‘occupation’ of the major groove at the centre of the modified helix by the two C_L–G_L central base pairs, which is proper to Z-DNA. The view of the central portion of the molecule, presented in Figure 5, shows that, despite these modifications, the helix is maintained all along. Note, for instance, the reversion of the C_L5 sugar relative to the G4 sugar, and that of the G_L6 sugar relative to the C_L5 sugar; the main direction of the chain, relative to G4, is therefore restored by the second sugar reversion. Also remarkable are the large overlaps occurring between the bases of C_L5 and C_L15 (interstrand) and the sugar of C_L5(15) and the base of G_L6(16) (intrastrand), leading to strong stacking interactions.

Figure 4.

Views of (a) the structure of d(CG)₅ calculated with the torsion angle dynamics program using synthetic data (an orange ribbon represents the backbone) and (b) the average structure of the l-modified decanucleotide calculated with experimental data.

Figure 5.

View of the four central base pairs of the average structure of the l-modified decanucleotide.

All these features are typical of a Z*-DNA conformation, which accounts for the NMR data, including the particular nOes and the chemical shifts. For example, the position of the H5′ proton of C_L5(15) is just above the six-membered ring of G_L6(16), which explains its high field shift.

Particular stacking interactions in the central portion

Figure 6 provides a view of the possible stacking interactions occurring within the two central C_L–G_L base pairs: (a) B-DNA, (b) Z-DNA, (c) l-modified decanucleotide and (d) the mirror image Z*-DNA. There is no doubt that the stacking pattern of the modified oligomer is clearly of the Z* type.

Figure 6.

Top views showing the stacking interactions of the CpG motif in (a) B-DNA, (b) Z-DNA, (c) l-modified decanucleotide and (d) mirror image Z-DNA.

Figure 7 shows the stacking of the central C_L–G_L base pairs (in red) with their neighbouring ^mC₇–G₁₄ and ^mC₁₄–G₄ base pairs (in green) under different conformations: (a) the B-DNA, (b) the modified decanucleotide, and (c) the mirror-image Z*-DNA. In the modified DNA (b) and in the standard mirror image Z*-DNA (c), the central guanines are pushed away from the centre to the periphery of the helix. The cytosine ring belonging to the neighbouring base pairs (^mC7–G14 and ^mC17–G4) is stacked onto these peripheral guanines. Thus, in both (b) and (c), the central C_LpG_L step in the Z-DNA structure facilitates the stacking with the immediately preceding and succeeding steps, both of which exhibit some Z-type features. It can be further noted that the stacking of cytosines with guanines at the centre of the helix is better achieved with the l enantiomer than with the d enantiomer.

Figure 7.

Top views of the four central bases in (a) B-DNA, (b) l-modified decanucleotide and (c) mirror image Z-DNA (the central CpG step is in red and the two neighbouring base pairs are in green).

The right-handed BZ*B helix

Our results prove, without any ambiguity, that the central C_LpG_L motif adopts a right-handed Z*-type DNA conformation. This also spans the d-deoxynucleotides that precede and succeed this motif. The originality of the l-modified helix resides in the larger base stacking at its centre compared to the classical left-handed Z-DNA. Indeed, the introduction of an l-deoxynucleotide in a chain of d-deoxynucleotides creates an important structural problem for the double-helix. If the l-sugar points in the same direction as that of the preceding d-sugar, the sense of the helix will be reversed. The major consequence will be a significant loss of base stacking and relative to the major (respectively minor) groove side of the preceding d-deoxyribonucleotide, the l-deoxynucleotide will present its minor (respectively major) groove side. Note that C_L5(15) in the modified decanucleotide is constrained to a sugar reversion that helps to maintain base stacking with its preceding d-deoxynucleotide G4(14). The base stacking between the G_L6(16) and its succeeding C7(17) is also preserved, but this requires the passage of the guanine in the syn conformation. As a result, the minor and major groove sides of the l-deoxynucleotide bases are not inverted with respect to the l-deoxynucleotide bases. Furthermore, the O4′ atom of the C_L5(15) l-sugar is positioned directly above the G_L6(16) six-membered ring. Such a system is stabilised via two types of interaction: an intracytidine O4′…H6–C6 hydrogen bond and an n→Π* interaction with the guanidinium ring (30).

Within the central C_LpG_L motif, the guanine prefers to rotate its base and adopt the syn conformation while the cytosine inverts its sugar, which again results in an inversion of the base orientation. At first sight, it may appear intriguing that, in response to stereochemical modifications, the central portion of the molecule adopts a conformation of Z*-DNA type rather than B*-DNA type. Yet, compared with the structural arrangement of the B*-DNA (Fig. 8), the Z*-DNA permits the best stacking at every step, including the steps at the junctions. Our study suggests that a right-handed helix, integrating just two B–Z* junctions, is stable if a good base–base/sugar stacking is achieved. We do not know, however, whether it is the number of l-deoxynucleotides versus the number of d-deoxynucleotides that determines the handedness of the double-helix. We presume that a larger number of l-deoxyribonucleotides would entail the passage to a left-handed double-helix.

Figure 8.

Stacking of one C_L–G_L base pair (in red) with its preceding G–^mC base-pair (in green) in an all B-DNA structure.

Biological implications

It has been repeatedly demonstrated that DNA fragments constructed from l-nucleotides, despite their better stability towards nucleases, cannot be used as pharmaceutical tools in the antisense strategy (3,4). The main reason invoked was the difficulty for l-DNA strands that resolve into left-handed helical structures to associate to natural DNA or RNA targets. The present results show that l-nucleotides containing DNA chains may assume a right-handed helical structure. This could allow a better strand adjustment to complementary d-DNA or RNA targets, even if good base-pairing requires some accommodation of partners. DNA strands exhibiting BZ*B type conformation could therefore conceivably be used in ‘gene’ therapy (4). Of course, we still need physicochemical and thermodynamical studies of such strands complexed to their natural targets. We believe also that the possible production of DNA-antibodies by Z*-DNA fragments deserves investigation. The N7 atom of the guanine aromatic ring, probably responsible for the production of antibodies against the natural left-handed Z-DNA, is exposed at the periphery of the double-helix in both left-handed Z-DNA and right-handed Z*-DNA and inversion of the helix handedness from left to right could in either case modulate the antigenicity. In addition, single-stranded decanucleotides containing a CpG motif are immunostimulatory (31,32) although the mechanism by which cells detect CpG is not yet totally established. A comparative analysis of a BZ*B strand versus a BB*B strand could offer a good opportunity to test the influence of the CpG chirality on the above-mentioned effect.

In conclusion, the C_LpG_L step at the centre of a l-deoxynucleotide chain may adopt the rare Z-DNA form at physiological salt concentrations. The chimeric molecule apparently resorts to the Z-DNA conformation to solve the problem of stacking continuity along the resulting double-helix.

SUPPLEMENTARY MATERIAL

Chemicals shifts, the constraints file and the ¹³C spectra are available as Supplementary Material at NAR Online.

PDB no. 1FV7