The 5′-O-DMT-2′-O-MOE-2-thiothymidine-3′-phosphoramidite was prepared according to the published procedure (11). The oligonucleotide d(GCGTA)-m⁵s²Umoe-d(ACGC) was synthesized on functionalized controlled pore glass (CPG) on an automated solid phase DNA synthesizer with final DMT group retained at the 5′ end. Standard phosphoramidites and solid supports were used for incorporation of A, T, G and C residues. A 0.1 M solution of the amidites in anhydrous acetonitrile was used for the synthesis of the modified oligonucleotide. For incorporation of 2-thio-2′-O-MOE-5-methyl U amidite, 6 equivalents of phosphoramidite solution were delivered in two portions, each followed by a 5 min coupling wait time. All other steps in the protocol supplied by the manufacturer were used without modification. Oxidation of the internucleotide phosphite to the phosphate was carried out using tert-butyl hydroperoxide/acetonitrile/water (10:87:3) with 10 min oxidation wait time. The stepwise coupling efficiencies were >97%. After completion of the synthesis, the solid support was suspended in aqueous ammonium hydroxide (30 wt%) and kept at room temperature for 2 h. The solid support was filtered and the filtrate was heated at 55°C for 6 h to complete the removal of all protecting groups. The crude oligonucleotide was purified by high performance liquid chromatography (HPLC, C-4 column, Waters, 7.8 × 300 mm, 15 µm, 300 Å, buffer A = 100 mM ammonium acetate, pH 6.5–7, buffer B = acetonitrile, 5–60% of B in 55 min, flow 2.5 ml min⁻¹, λ 260 nm). Detritylation was achieved by adjusting the pH of the solution to 3.8 with acetic acid and keeping at room temperature until complete removal of the trityl group, as monitored by HPLC analysis. The oligonucleotide was then desalted by HPLC and characterized by ES-MS (calculated mass: 3118.35 g mol⁻¹; found: 3119.57 g mol⁻¹) and purity (91% full length) was assessed by capillary gel electrophoresis.

Crystallization trials were performed with the Nucleic Acid Mini-screen (24) by Hampton Research (Aliso Viejo, CA), using the hanging drop vapor diffusion technique. Droplets with volume 2 µl of a 1:1 mixture of sample and mini-screen buffer were equilibrated against 1 ml of 35% 2-methyl-2,4-pentanediol (MPD). Crystals were obtained from a droplet that contained 1 mM oligonucleotide, 40 mM sodium cacodylate pH 7.0, 12 mM spermine tetrahydrochloride, 12 mM sodium chloride, 80 mM potassium chloride and 10% v/v MPD. A single crystal was mounted in a nylon loop and frozen in liquid nitrogen. Diffraction data were collected at low temperature in a cold nitrogen stream on the 5-ID beamline of the DuPont-Northwestern-Dow Collaborative Access Team at the Advanced Photon Source, Argonne, IL. Separate data sets for high and low resolution reflections were acquired. All data were processed with the program XDS (25) and a summary of selected crystal data and data collection statistics is listed in Table 1.

Due to the close similarity between two of the unit cell constants, the diffraction data were processed both in the space groups P222 (orthorhombic) and P422 (tetragonal). The resulting values for R-merge differed only minimally (<1%) and structure determination was subsequently attempted in P2₁2₁2₁, the space group found for the vast majority of crystals of A-form DNA 10mers with isolated 2′-O-modified residues investigated in our laboratory (20), and in space group P4₂2₁2. Phasing was carried out by the molecular replacement method using the program AMORE (26) as part of the CCP4 suite of crystallographic programs (27). An A-form DNA decamer of the same sequence GCGTATAGCG (28) was used as the search model. Promising solutions with good R-factors, reasonable correlation coefficients, and few short lattice contacts were found in both space groups. Initial refinements of the models were performed with the program CNS (29), setting aside 5% randomly selected reflections for calculating the R-free (30). While the R-factor calculated for the model in the orthorhombic space group P2₁2₁2₁ dropped rapidly, it failed to drop below 35% in space group P4₂2₁2. Consequently, refinement was continued only in P2₁2₁2₁. After rigid-body refinement and simulated annealing plus multiple rounds of positional and individual B-factor refinement, the complete model, including the 2-thio and 2′-O-MOE modifications for the two modified residues, exhibited an R-factor of 30.1% (R-free of 28.5%). Refinement was then continued with the progam REFMAC (31), using a bulk solvent correction and treating DNA atoms and water molecules with restraint anisotopic B-factors (32,33). The final model refined to an R-factor of 19.4%, using all reflections up to a resolution of 1.61 Å. Selected refinement parameters are listed in Table 1 and an example of the quality of the final Fourier sum (2F_o-F_c) electron density is depicted in Figure 3.

Final coordinates and structure factors have been deposited in the Protein Data Bank () PDB ID code 2AXB.

Owing to the similar b and c unit cell constants (Table 1), diffraction data from a single crystal of the m⁵s²Umoe-modified DNA decamer were indexed and processed both in orthorhombic and tetragonal space groups. Following structure determination with molecular replacement using an A-form model, initial rounds of refinement in the space groups P2₁2₁2₁ and P4₂2₁2 indicated that the crystal system is orthorhombic. Thus, the crystallographic asymmetric unit consists of a single duplex and the lattice features a non-crystallographic 4-fold symmetry. Multiple rounds of coordinate and temperature factor refinements and simulated annealing led to an improved model, and electron density maps based on it revealed ordered 2′-O-MOE substituents in the minor groove. The positions of sulfur atoms were marked by spherical peaks of density that are clearly enlarged relative to those around the exocyclic O4 and O2 atoms of unmodified thymines (Figure 3). In the duplex, residues of one strand are numbered from 1 to 10, and those in the complementary strand are numbered from 11 to 20; m⁵s²Umoe residues are located at positions 6 and 16.

The modified duplex adopts an A-form conformation with an average helical rise of 2.64 Å (SD 0.69 Å) an average helical twist of 34.3° (SD 3.4°). The average inclination of base pairs relative to the helical axis is 15.6°. Except for residues at the 5′-terminal ends, the conformations of all sugars fall into the C3′-endo range. The deoxyriboses of residues G1 and C2 adopt C2′-exo and C4′-exo puckers, respectively, and G11 is flipped into the C2′-endo conformation. Similarly, most of the backbone torsion angle conformations observed in the decamer duplex are those typically associated with an A-form geometry (-sc/ap/sc/sc/ap/-sc; α to ζ). Exceptions are constituted by A5 in strand 1 and G13 in strand 2 that both assume an extended backbone variant with α, β and γ in the ap conformation. The conformation is brought about by a crankshaft motion of the backbone and is accompanied by elongated distances between adjacent intra-strand phosphorus atoms (6.9 Å between A5 and m⁵s²Umoe6 and 6.4 Å between G13 and T14; avg P^…P distance: 6.1 Å).

To analyze the geometry of the m⁵s²Umoe-modified duplex further and to determine potential changes as a result of the presence of sulfur atoms in the central A:m⁵s²Umoe base pairs, we used the previously determined structure of the duplex with identical sequence and m⁵Umoe residues at positions 6 and 16 as a reference (20). The resolution of that structure is similar (1.93 Å) to that of the m⁵s²Umoe-modified duplex described here. Both crystals belong to space group P2₁2₁2₁ and have very similar cell constants [a = 24.93, b = 44.59, c = 45.38 Å (reference) versus a = 25.25, b = 45.00, c = 44.97 Å]. We will refer to the reference and the 2-thio-modified duplexes as the m⁵Umoe and m⁵s²Umoe structures, respectively.

Because the conformational perturbations caused by the 2-thio modification can be expected to be quite subtle, it is important to identify and quantify conformational differences between the m⁵s²Umoe-modified and reference duplexes that are due to crystal packing effects. Interactions between neighboring duplexes in the crystal lattices of m⁵Umoe and m⁵s²Umoe decamers involve stacking of terminal base pairs into the outer portions of the minor groove (34). This particular mode of interaction triggers various degrees of kinking into the major groove. Individual crystallization conditions, the chemistry and spatial properties of substituents of 2′-O-modified residues and their location in the decamer strand, as well as data collection temperature all appear to play a role in the relative degree of kinking (35). Superimposition of the two duplexes reveals that the m⁵Umoe decamer exhibits a slightly stronger kink than the m⁵s²Umoe decamer (Figure 4). The kink compresses the major groove and occurs between the tetramer 5′-G1pC2pG3pT4 and the hexamer pA5p(m⁵Umoe6)pA7pC8pG9pC10. It is accompanied by significant changes in the local base-pair (stagger), base-pair step (rise, twist and roll) and base-pair helical parameters (inclination). The absolute values of these parameters at that site constitute either the minimum (stagger [m⁵Umoe], rise, twist) or maximum (stagger [m⁵s²Umoe], roll, inclination) among the individual observations with both duplexes. Most probably, these changes are only weakly affected by the oxygen→sulfur substitution in residue 16, although the kink occurs between the T4:A17 and A5:m⁵s²Umoe16 pairs. Rather, they are a reflection of the intrinsic conformational properties (‘bendability’) of the decamer sequence and the packing forces in the orthorhombic crystal lattice. Regarding the possible origins of the observed difference in kinking in the two duplexes, it is noteworthy that the m⁵Umoe structure was determined at room temperature (20) whereas the m⁵s²Umoe structure is derived from diffraction data collected at −170°C. Thus, slightly divergent packing forces due to the temperature change are the most likely cause of the observed differences between the global conformations of the m⁵s²Umoe and m⁵Umoe duplexes. Because of the kink adjacent to the A5:m⁵s²Umoe16 base pair and the ensuing perturbations in its immediate vicinity, one would expect that careful analysis of the geometry of the other modified base pair, m⁵s²Umoe6:A15, might yield a more reliable estimate of the changes as a result of the 2-thio modification.

Both structures were analyzed with the program 3DNA that allows calculation of geometric parameters for nucleic acid molecules (36). Among the parameters, one would expect those that concern the local geometry of base pairs and the stacking interactions to be the most informative in terms of potential conformational changes due to 2-thio modification. Figure 5 depicts graphic representations of selected geometric parameters for base pairs in the m⁵s²Umoe and reference duplexes. A significant change (11°) is observed in the parameter opening for base pair m⁵s²Umoe6:A15 compared with the corresponding pair in the reference structure (Figure 5A). This change in opening results in a larger separation between the C2 atoms of the two bases in the minor groove with the 2-thio modified pair [4.23 Å versus 3.89 Å (ref. structure); Figure 6A]. Comparison of the stretch parameters also indicates a small change (0.15 Å; not shown) for this base pair that is related to accommodating the larger sulfur atom. Compared with the above change in opening, the differences in the parameters shear, stagger and propeller twist (Figure 5B–D, respectively) are relatively minor for base pair m⁵s²Umoe6:A15. Likewise, it does not appear that buckling of bases is affected by the sulfur substitution. The calculated values for both the m⁵s²Umoe6:A15 and the A5:m⁵s²Umoe16 pairs differ by <1° from those for the corresponding pairs in the reference structure. Interestingly, the latter base pair does not exhibit an altered opening angle relative to the reference duplex (Figure 5A). The difference amounts to less, ∼2° although the separation between C2 atoms of bases (4.11 Å) is comparable with that seen in the m⁵s²Umoe6:A15 pair (see above). Similarly, other parameters, including stretch, shear, stagger and propeller twist change subtly, but none of them appears to be making the major contribution such as opening in case of the m⁵s²Umoe6:A15 pair (Figure 5). Rather, the combination of many small changes results in an arrangement that shifts the sulfur atom away from the C2 atom of adenine in the minor groove. In any case, a superposition of the s²m⁵Umoe6:A15 and A5:s²m⁵Umoe16 pairs demonstrates that they adopt relatively similar conformations, with virtually identical relative orientations of C2(A) and S2(m⁵s²Umoe) (Figure 6B).

In order to examine whether presence of the sulfur atoms in the m⁵s²Umoe structure led to changes in the stacking interactions involving the four central base pairs, we calculated the overlap areas between adjacent pairs in Ų units (36). The 3DNA program allows computation of the surface areas of polygons projected in the mean plane of a base pair at a particular step, including exocyclic ring atoms. Stacking interactions at selected base-pair steps are shown in Figure 7. Inspection of the plotted diagrams for the m⁵s²Umoe and reference duplexes reveals minor alterations in stacking at most base-pair steps. An exception is found at the (A5)p(m⁵s²Umoe6):(A15)p(m⁵s²Umoe16) step (Figure 7C) that displays a clearly visible shift between base pairs relative to the reference structure. Indeed, the calculated y-displacement at that site shows a difference of 2.3 Å between the m⁵s²Umoe and m⁵Umoe structures (1.1 versus −1.2 Å, respectively; Figure 5E). The difference in x-displacement is somewhat smaller, −1.3 Å (−5.3 versus −4.0 Å in m⁵s²Umoe and m⁵Umoe, respectively; Figure 5F), but constitutes the largest change for that parameter in the entire duplex. The two neighboring steps also show considerable changes in y-displacement that amount to −1.6 Å (Figure 7B) and −1.7 Å (Figure 7D). At the central step, it is readily apparent that the two sulfur atoms have moved toward one another. Indeed, the distance between the sulfur atoms (3.48 Å) is shorter than the one between the corresponding O2 atoms in the reference structure (3.73 Å). The former is slightly shorter than the sum of van der Waals radii for the two sulfur atoms (3.6 Å). When the overlap areas between adjacent base pairs at the (T4)p(A5):(m⁵s²Umoe16)p(A17), (A5)p(m⁵s²Umoe6):(A15)p(m⁵s²Umoe16) and (m⁵s²Umoe6) p(A7):(T14)p(A15) steps (Figure 7B–D) are added, a slight increase in stacking for the m⁵s²Umoe decamer compared with the reference structure is apparent (17 Ų versus 15 Ų). Therefore, we can conclude that the experimental data regarding stacking are at least in line with the stabilizing effect of the 2-thio modification.

In the minor groove, sulfur atoms of adjacent m⁵s²Umoe residues are bridged by a single water molecule. This water, in turn, forms a hydrogen bond to the outer oxygen of the 2′-O-MOE substituent of m⁵s²Umoe16, and a second, relatively weak hydrogen bond to the 4′-oxygen of the same residue. The 2-sulfur atom of m⁵s²Umoe6 is hydrogen bonded to a second water molecule. This water bridges the nucleobase to a further water molecule that is itself hydrogen bonded to both O2′ and the outer oxygen of the 2′-O-MOE substituent from m⁵s²Umoe6. Both 2′-O-MOE moieties adopt gauche conformations (m⁵s²Umoe6, sc⁺, and m⁵s²Umoe16, sc⁻, Figure 3). In the case of m⁵s²Umoe6, the substituent is directed toward the center of the minor groove, allowing for a direct link between the MOE moiety and the minor groove edge of the base by a single water molecule. On comparison, the substituent of m⁵s²Umoe16 is directed away from the base edge, therefore requiring a water tandem to link MOE moiety and S2. Water molecules trapped between the 2′- and the outer oxygen atoms of MOE substituents are connected to phosphate groups from 3′-adjacent residues via single- or double-water bridges, a particular feature previously observed in the crystal structures of 2′-O-MOE-modified RNAs (20,23) and shared by other modifications with 2′-O-substituents containing hydrogen-bond acceptors and/or donors (37). The presence of a stable network of water molecules in the minor groove that connects m⁵s²U bases with sugar moieties and phosphate groups provides an indication that the 2-thio modification does not significantly alter the water structure compared with uracil. This observation alone does not allow conclusions as to the role of hydration in the overall stabilizing effect of the 2-thio modification. Nevertheless, the absence of obvious disruptions in the minor groove hydration by sulfur atoms in the m⁵s²Umoe duplex structure clearly argues against a destabilizing contribution.