The Conformationally Constrained N-Methanocarba-dT Analog Adopts an Unexpected C4'-exo Sugar Pucker in the Structure of a DNA Hairpin

Abstract

Incorporation of a bicyclo[3.1.0]hexane scaffold into the nucleoside sugar was devised to lock the embedded cyclopentane ring in conformations that mimic the furanose North and South sugar puckers. To analyze the effects of North-methanocarba-2'-deoxythymidine (N-MCdT) on the conformation of B-DNA, we crystallized d(CGCGAA[mcTmcT]CGCG) with two N-MCdTs. Instead of a duplex the 12mer forms a tetraloop-hairpin, whereby loop N-MCdTs adopt the C4'-exo pucker (NE; P=50°). This indicates that the bicyclic framework does not limit the sugar pucker to the anticipated C2'-exo range (NNW; P=−18°).

Conformationally restricted or locked nucleoside analogs are being investigated in biochemistry, biotechnology and medicinal chemistry and are of considerable interest in nucleoside-based drug discovery and, incorporated into oligonucleotides, for use as antisense agents, siRNAs and aptamers.^1–8 There are multiple strategies to constrain the furanose sugar in the North conformation (i.e. C3'-endo; Figure 1). 2'-Carbohydrate modifications,^9,10 including locked nucleic acid (LNA),¹¹ and analogs with the 3'-oxygen replaced by a less electronegative moiety, i.e. nitrogen as in N3'→P5' phosphoramidate DNA,^12,13 all limit the sugar pucker to the northern range to various degrees. A large number of such analogs were explored over the last 20 years owing to their increased RNA affinity as a result of preorganizing the modified strand for the A-type conformation.^14,15 Although conformational preoganization is often assumed to be synonymous with an entropically favourable effect, the higher RNA affinity typically observed for analogs with N-sugars has been shown in some cases to be due to enthalpic gains instead.^16,17

Figure 1.

Pseudorotation cycle of the five-membered sugar ring in nucleosides with values of phase angles indicated in multiples of 36°.¹⁸ Selected pucker modes are labeled at the periphery and illustrated with representative examples of sugars adopting a particular conformation: C2′-exoNorth-bicyclo[3.1.0]hexane (N-MCdT; magenta), C3′-endo 2′-deoxyribose (A-form duplex, similar for ribose; black), C4′-exo MCdT (this work; magenta), O4′-endo 2′-fluoro-2′-deoxyarabinose (FANA; light green), C1′-exo arabinose (ANA; green); C2′-endo 2′-deoxyribose (B-form duplex; black), and C3′-exoSouth-bicyclo[3.1.0]hexane (S-MCdT; blue).

Compared with chemical modifications that constrain the furanose in the N range, those maintaining a Southern (S) pucker (i.e. C2'-endo, Figure 1) are much less common. Remarkably, the bicyclo[3.1.0]hexane scaffold that contains a cyclopentane ring provides a surrogate of the five-membered furanose ring in nucleosides, whereby the cyclopentane is locked in an envelope conformation (Figure 1).¹ The location of the fused cyclopropane ring then determines whether the pseudoboat conformation adopted by bicyclo[3.1.0]hexane mimics N (C2'-exo, C4'-C6'-C7' cyclopropane; Figure 1, red) or S (C3'-exo, C1'-C6'-C7' cyclopropane; Figure 1, blue) puckers of furanose.

The corresponding N- and S-2'-deoxy-methanocarba-nucleosides (N- and S-MCdNs, respectively) or nucleotides have been tested for their antiviral activities and in SAR studies involving kinases, polymerases and reverse transcriptases.^1,19–22 The effects of such nucleotides incorporated into DNA oligo-nucleotides²³ on the pairing stability and duplex conformation (i.e. bending) have also been reported.^24–27 Despite the extensive biochemical and biophysical evaluation of methanocarba-nucleotides and -modified oligonucleotides, experimental, atomic-resolution models of N- or S-MCdNs in the oligonucleotide context and their effects on DNA conformation are presently lacking.

To establish the preferred pucker of N-2'-deoxy-methanocarba nucleotides, we crystallized the Dickerson-Drew dodecamer (DDD) d(CGCGAAmcTmcTCGCG) with two N-MCdTs. The structure was determined by single-wavelength anomalous dispersion (SAD) using two Ba²⁺ sites and refined to 1.8 Å resolution (see the supporting information for details).

An example of the quality of the final electron density is shown in Figure 2. The volume of the crystallographic unit cell was consistent with a single duplex per asymmetric unit. However, to our surprise the experimental electron density revealed two independent single-stranded dodecamers, each folded into a hairpin with a left-handed [d(CGCG)]₂ stem and an AAmcTmcT loop (Figure 3). To our knowledge this is the only stem-loop structure for the DDD reported to date.

Figure 2.

(A) Final Fourier (2Fo-Fc) sum electron density (1s threshold) around the sugar of N-MCdT7 (C4′-exo, P = 50°). For comparison, 2′-deoxyriboses adopting (B) C2′-exo (P = −18°) and (C) C3′-endo (P = +18°) puckers are shown (see also the SI).

Figure 3.

Hairpin structure of the N-MCdT-modified DDD. (A) DNA sequence, nucleotide numbering and head-to-tail arrangement of left-handed [d(CGCG)]₂ stems. (B) The two independent hairpin molecules with residue coloring matching that in panel A. (C) Superimposition of the two independent hairpins reveals their nearly identical conformations. (D) Close-up view of the conformation adopted by the AAmcTmcT loop.

Z-DNA is preferentially formed by alternating CG-sequences at high salt, whereby G and C exhibit the typical C3'-endo/syn and C2'-endo/anti conformations, respectively.²⁸ At the transition between stems from adjacent hairpins oriented in head-to-tail fashion, the virtual absence of a twist and a large slide between C:G pairs result in stacked guanines (Figure 3B). Both hairpins adopt virtually identical structures and AAmcTmcT loops can be neatly overlaid (Figure 3C). The loop arrangement is stabilized by a partial stacking interaction between G4 and A5 and a continuous A6, mcT7, mcT8 and C9 stack (Figure 3D). Other stabilizing influences include C-H···O hydrogen bonds between C8-H [A5] and O4 [T7] as well as C5M (methyl) and O2P (T7; intra-nucleotide). Most noteworthy in regard to our original goal of establishing the preferred conformation of the N-MCdT sugar is the C4'-exo conformation of both mcT7 and mcT8 (Figures 2A, 3D). The pseudorotation phase angles as calculated with PROSIT²⁹ are 50.1° (mcT7) and 52.5° (mcT8). The corresponding angles in the second hairpin are 48.5° and 51.5°, respectively. Therefore, instead of the expected C2'-exo pucker, N-methanocarba-nucleotides adopt a Northeast conformation in the present structure (Figure 1).

In the crystal hairpins form infinite columns; stems are stacked at one end of each hairpin (Figure 3B), and at the other, loops are interacting under formation of three base pairs (Figure 4). Nucleotides mcT7 from joined hairpins form a sheared T:T^# pair in the center (# marks a nucleotide from a second hairpin), stabilized by two N3-H···O2 hydrogen bonds. This pair is flanked by Watson-Crick type pairs between A6^# and mcT8 and A5 engages in a Hoogsteen pair with the former via N6-H···O2 hydrogen bonds, thus resulting in base triples (Figure 4).

Figure 4.

Cross-eye stereo view of the interactions between hairpin loops in the crystal structure of the N-MCdT-modified DDD. Carbon atoms of residues G4, A5, A6 and C9 in hairpin 1 are cyan and in hairpin 2, the corresponding atoms are gray. Hydrogen bonds are shown as thin solid lines and all water molecules were omitted for clarity.

A search for hairpin structures that would allow a comparison with the DDD hairpin studied here did not yield cases of DNA AATT³⁰ or RNA AAUU³¹ tetra-loop sequences. Many years ago, Dickerson and coworkers determined the crystal structure of the DNA hexadecamer d(CGCGCGTTTTCGCGCG).^32,33 The oligonucleotide forms a hairpin with a left-handed [d(CGCGCG)]₂ stem and a dT₄ loop. However, the arrangement of thymidines in their structure is completely different from that of the AAmcTmcT loop observed here and thus does not allow a meaningful comparison or conclusions regarding the preferred sugar pucker of N-MCdT residues.

The DDD with two incorporated N-MCdT nucleotides crystallized as a hairpin, which is highly unusual. At the concentrations of between 0.1 and 1 mM typically used for crystallization of an oligonucleotide, formation of a bimolecular species (i.e. the duplex) is the rule. Crystallization experiments with RNA hairpins featuring tetrameric stems and tetra-loops invariably lead to duplex formation, with or without mismatch base pairs (see, for example, ref. 34). Only unusually stable constructs with short stems, i.e. those capped with trans-stilbene,^35,36 crystallize as hairpins. Because the structure of the chemically modified DDD described here represents the only case of such a dodecamer adopting a stem-loop conformation in the crystal, the question arises as to the origins of this behavior.

Even DDDs containing 2'-modified T or U residues at positions 7 and 8, i.e. 2'-SMe-U,³⁷ that display a C3'-endo pucker, crystallized as duplexes with enlarged minor grooves and minimal bending, but not as hairpins. Moreover, neither NMR nor CD experiments provided any indication for the presence of a hairpin in solution.^24–26 Modeling of methoxy groups at C2' of N-MCdT7 and -T8 results in several short distances between 2'-oxygen and methyl carbon and backbone and base atoms (Figure 5). Therefore, it is unlikely that a DDD with rU(T) or 2'-OMe-rU(T) nucleotides at positions 7 and 8 would adopt the hairpin conformation observed here for the N-MCdT-modified dodecamer. Interestingly, the chimeric DDD 5'-fCfGfCfGaAaAaUaUfCfGfCfG-3' (fN=FANA, aN=ANA; Figure 1) was found to exist in an equilibrium between duplex and a hairpin with an aAaAaUaU loop.³⁸

Figure 5.

The shortest distances involving 2′-O-methyl groups (methyl carbons highlighted as yellow spheres) modelled onto the crystal structure of the N-MCdT-modified DDD hairpin. Only a portion of the loop, mcT7/8, and the adjacent stem, C9, is shown.

The sugar puckers of N-MCdTs in the crystal structure differ by 70° in terms of phase angle from the value predicted for¹ and typically associated with this modified nucleotide (Figure 1).^24–27 At the nucleoside level, adoption of the C4'-exo pucker likely results in a higher potential energy compared with the C2'-exo pucker (see SI for details). Although the structure leaves us without a model of the conformation of the N-MCdT sugar inside a DNA duplex, we have to conclude that the adoption of the hairpin by the modified DDD is linked to the presence of methanocarba-residues. At the very least, we have identified an alternative, stable NE conformation for the bicyclic nucleoside analog that is allegedly constrained to the N pucker region. Alternatively, the observed sugar pucker represents its actual preference in an oligonucleotide context and the nucleoside should be termed NE-MCdT instead of N-MCdT.